CEO Collect Earth Online

Using Interpretation Keys with CEO

Using Interpretation Keys with CEO

Imagine that you are leading a team of interpreters. Your team’s job is to interpret hundreds of plots in Collect Earth Online (CEO) in order to gather data about land cover using remotely sensed imagery and a list of mutually exclusive land cover types of interest.

To successfully accomplish this goal, each interpreter must accurately and consistently identify each land cover type. That is, forests must always be identified as forests, wetland areas as wetland areas, and so on.

How do you make sure that your team can successfully accomplish this task?

Interpretation keys, also called photo interpretation keys, offer a powerful tool to create consensus, documentation, and institutional knowledge. Interpretation keys are used by teams around the world in order to successfully collect data in CEO.

But what are interpretation keys? Why should you create one, and how do you use them? Read on to find out, in this first blog post on land cover and land use interpretation using CEO.

🤔 What are interpretation keys?

Interpretation keys provide interpreters with a guide for how to examine remotely sourced imagery, and how to classify land cover and land use as well as specific events based on different ‘signatures’ of the imagery. These signatures include: location, size, shape, shadow, tone and color, texture, pattern, height or depth, and situation or context. They can include both quantitative information (e.g. the typical wavelengths reflected by that land cover) and qualitative information (e.g. roads near a cleared forest may indicate logging). An interpretation key will provide this information for each land cover, land use, or any specific events such as fire, landslides, or logging that is of interest to the project.

An interpretation key also usually includes information about the specific project (or projects) for which it was developed, any project-specific land cover and land use definitions, any standards (e.g. IPCC) that the project follows, and records information generated by the team about how to reach a decision when the image is ambiguous. It may also include information about which imagery should be used, from what time of year the imagery should come, and other useful information.

Projects like this one in CEO benefit from having robust interpretation keys developed prior to data collection efforts.
Projects like this one in CEO benefit from having robust interpretation keys developed prior to data collection efforts.


If you’d like to dive more deeply into interpretation keys and their history, there are some great resources available online.

🌟 Why create an interpretation key?

Being able to identify landscapes and landscape changes using remote sensing data and time-series information is an important skill for creating training data, verifying algorithm outputs, and creating sample-based estimates of area.

Creating an interpretation key is important to support all of these tasks and serves multiple purposes, including:

  • Creating consensus. The interpretation key helps your team build a shared understanding of what each land cover type is and how to identify it. This means that if you have multiple interpreters, they should be able to classify land cover categories in the same way.
  • Creating documentation. The key records what imagery fits your label definitions for your project. This is important for funding and publishing.
  • Creating institutional knowledge. An interpretation key allows new team members to understand what existing team members consider to be defining characteristics of each land cover type. This helps your new team members start collecting data and contributing to your project quickly and accurately.

👩‍🏭 Putting it all together

But what does this mean practically?

We can examine an existing interpretation key to better understand what they are. For example, the TerraBio team recently conducted a data collection project using CEO for which they created a forest change interpretation key.

Example pages from the Best Interpretation Practices section detailing how to use the available imagery to accurately answer the survey questions in CEO.
Example pages from the Best Interpretation Practices section detailing how to use the available imagery to accurately answer the survey questions in CEO.


Let’s examine the major sections used in this interpretation key and what types of information they contain:

  1. Background on the study region. Information about the area where the data collection is taking place, including the land use history.
  2. Background on the types of forest change events of interest. Information about forest degradation, deforestation, and natural regeneration and reforestation results.
  3. Best interpretation practices. Step-by-step instructions on what imagery to look at and for which years in order to answer the CEO survey questions accurately.
  4. Example Imagery and Indexes. This is the core of the interpretation key, containing multiple examples for each of the types of forest change of interest. For this project, the forest changes of interest detailed in the key included Forest Degradation, Deforestation, and Regeneration and the events causing the change included Mining, Transition to Pasture, Fragmentation, Fire, Clear Cutting, and Selective Logging.
  5. Further Reading and Resources. Resources that interpreters can use to learn more about the project, the land cover, land use, or change events being detected.
Example from the Interpretation Key: Deforestation section of the interpretation key highlighting multiple examples of deforestation taking place. This example highlights what the Forest to Pasture transformation looks like.
Example from the Interpretation Key: Deforestation section of the interpretation key highlighting multiple examples of deforestation taking place. This example highlights what the Forest to Pasture transformation looks like.


In future blog posts, we will explore how to create an interpretation key, and how to identify certain land covers, land uses, and landscape change events.

CEO would like to thank its ongoing funders FAO, NASA–USAID SERVIR, and SilvaCarbon, a US government program. Thanks also to CEO’s technology partners: Norway’s International Climate & Forests Initiative for funding open high-resolution data availability; Planet for providing high-resolution imagery; and the Google Earth Engine team for creating a platform for Earth science data and analysis.

Collect Earth Online is working constantly to improve the user experience, and your feedback is invaluable. If you have ideas to share, please write to support@collect.earth.

Thank you!