Forest Typology 2020: Global 10-Meter Land Cover Classification

Global land cover data for 2020 at 10-meter spatial resolution, categorizing terrestrial areas into six forest and land-use types. The dataset was developed by Maxim Neumann using a deep learning pipeline trained on over 1.7 million globally distributed samples. It is aligned with FAO and EUDR definitions and distributed as Cloud Optimized GeoTIFFs.

Use Cases

• Monitoring deforestation and forest degradation based on the primary and naturally regenerating forest classes.
• Assessing compliance with regulations like the EU Deforestation-free Products Regulation (EUDR) using the aligned classification schema.
• Modeling carbon stocks and biodiversity by distinguishing between plantation, planted, and primary forest types.
• Analyzing agricultural and agroforestry landscapes using the 'Tree Crops and Agroforestry' and 'Other Land' categories.

Strengths

• High spatial resolution of 10 meters per pixel for detailed land cover analysis.
• Training on over 1.7 million globally distributed samples suggests a broad foundational dataset.
• Classification schema is explicitly aligned with international standards from the FAO and EUDR.
• Large dataset size of 23.9 GB indicates substantial geographic coverage and detail.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• The classification layer provides only hard labels (argmax), limiting uncertainty analysis without the companion probability dataset.

Provenance

Source

Maxim Neumann via figshare.

Collection Method

Developed using a three-stage cascaded deep learning pipeline.

Time Range

Baseline year 2020.

Freshness

Last updated 2026-05-22 23:22:02; freshness should be verified.

Geography

Global land areas between 65 degrees S and 84 degrees N latitude.