Google’s new forest map looks beyond green patches to show what kind of forest is there
On this page
Announcement date: October 22, 2026
A green patch on a satellite image does not tell the whole story. It could be a natural forest, a timber plantation or a farm growing tree crops. On October 22, 2026, Google DeepMind and its partners announced a global mapping project designed to make those differences easier to see.
The release, called Forest Typology 2020, or ForTy 2020, classifies land at a resolution of 10 metres. Google describes it as work with Google Research, the World Resources Institute and the International Institute for Applied Systems Analysis.
Why counting trees is not enough
Imagine two photographs taken from far above the ground. Both show rows of green. In one place, the trees form part of a varied natural habitat. In another, they are a crop managed for harvest. Calling both places “tree cover” leaves out a difference that matters to the people studying them.
A map needs categories that match the question being asked. If the question is simply whether trees are present, one broad label may help. If the question is how land is used, that label is too vague.
What was released
According to Google, the project sorts land into six forest and land-use types. Alongside the 2020 map, the team shared an early yearly series covering changes from 2017 to 2025. The aim is to show both forest change and what replaced a lost forest, such as a plantation or farmland.
The 2020 in the project’s name describes the map’s reference year, not the year of this announcement. That small distinction is important when a new research release uses older observations.
What does 10-metre resolution mean?
Think of a map as a grid. At this resolution, a square represents a patch of ground measuring about 10 metres on each side. A smaller grid can reveal more local detail than a very broad one, but it does not make every label certain.
Resolution and accuracy are different ideas. Resolution describes the size of the pieces. Accuracy describes how often the labels agree with reality. A detailed-looking map still needs careful evaluation.
A beginner’s way to read the results
Start with the legend, which explains the colours and labels. Check the reference year. Then ask what the map can reasonably tell you about a place you recognise. Do not assume it describes what is on the ground today if its observations come from an earlier period.
If two maps disagree, check whether they use the same definitions. One may count a plantation as forest while another separates it from natural forest. The disagreement may partly reflect different questions, not just a technical mistake.
This is a useful lesson beyond environmental research. In a spreadsheet, “active customer” might mean someone who bought this month or someone whose account is still open. A clear label matters as much as an attractive chart.
QueryFun’s take
This story shows AI being used for a task that is quite different from writing chat replies. Its value lies in helping researchers organise observations into meaningful categories. The responsible way to use such a map is to understand the definitions, date and uncertainty before drawing a conclusion.
Original source: Google DeepMind’s forest-typology announcement. Company statements are attributed to their source. Practical examples and commentary are QueryFun’s explanation, not an independent product test.