Open Forest Observatory Data Catalog

It is critical that you read this Description document in full prior to any use of the data. 

The Open Forest Observatory (OFO) Open Data Catalog consists of two components: (1) ground-based manual forest inventory plot data and (2) drone imagery and derived data (e.g. orthomosaics, canopy height models, and maps of automatically detected trees). Many of the ground plots are covered by one or more drone collects, facilitating development and validation of drone-based forest mapping methods.

The OFO data catalog is a compilation of contributions from many organizations and individuals. Each data item (e.g. ground-based manual inventory plot; drone collect and derived products) has its own Creative Commons data license, as indicated by the `license` attribute (column) in the associated data/metadata file(s) (see below).

The Zenodo data description (the document you are currently reading) describes the data catalog and important usage restrictions and considerations. It is critical that you read this Description document in full prior to any use of the data. Use of any data to produce any derivative products (or simply re-sharing the data as-is) requires (1) attribution to the data contributors (not the OFO) and specification of a data license equivalent to or compatible with the original data license, (2) acknowledgment of the Open Forest Observatory as the means used to discover and access the data, (3) inclusion of any metadata necessary for others to understand and use any OFO-derived data included in the derived data product, and (4) any other requirements specified in the license(s) of the datasets you use.

For questions around the use and interpretation of the data, contact openforestobservatory@ucdavis.edu.

Ground-based manual forest tree inventory plot data

These data files represent the locations of trees in plots that are mapped by technicians on the ground using traditional forest tree stem-mapping methods.

Data & metadata included in this Zenodo record:

• ofo_ground-reference_plots.gpkg: Plot-level data and metadata, including plot geospatial footprint, survey dates, protocol components such as tally tree definitions, and data license, in geospatial (geopackage) format

• ofo_ground-reference_trees.gpkg: Tree-level data, including tree stem geospatial location, species, size, in geospatial (geopackage) format

The OFO web catalog contains web preview functionality for ground reference plot data.

Important details and considerations

• The OFO ground reference data schema describes plot-level and tree-level attributes. It is critical that you read this document carefully, as it describes data attributes that could affect whether to include or exclude some plots depending on the analysis. For example, some plots consist of nested circles in which trees of different sizes were measured (large plot for large trees, small plot for small trees).

• Important usage restriction: We have designated a stratified subset of approximately 20% of plots to be excluded from training ML models that are intended for use by the forest mapping/ML community. For such applications, these plots are reserved for use as test data for model evaluation and intercomparison. The plots to be excluded from training are indicated in the plot table by the value true in the column withheld_from_training.

• As we build out the OFO public data catalog, we will add additional data tables, such as a “hyperplots” table that links nested plots of the sort described above. These data exist in the OFO internal database and can be provided upon request. The data schema linked above describes the OFO internal database, much (but not all) of which is replicated in the public data linked here.

• Survey protocol varies by plot. Protocols differ in many attributes, including what tree size attributes are measured (e.g., DBH, height, both), what trees are surveyed vs. ignored, etc.

• Data license varies by plot. All use Creative Commons licenses (most of them requiring data contributor attribution in derivative products), but please be aware that some plot data licenses include non-commercial and share-alike clauses that limit derivative works and their uses in certain contexts.

Drone-derived forest data

Metadata included in this Zenodo record:

• ofo_drone-missions_metadata.gpkg: Mission-level metadata, including geospatial footprint, flight altitude, raw image density, and data license, in geospatial (geopackage) format

Data files themselves are accessible via the Jetstream2 Object Store (S3-compatible bucket):

• Public HTTP URL template for orthomosaics (cloud-optimized geotiff): https://js2.jetstream-cloud.org:8001/ofo-public/drone/missions_03/<mission-id>/photogrammetry_03/full/<mission-id>_ortho-dsm-ptcloud.tif

• Many other data products (e.g., canopy height model, raw images) and previews/thumbnails are also available (see 'Data types', below. A representative (but truncated) bucket file listing illustrating file naming conventions can be obtained at: https://js2.jetstream-cloud.org:8001/swift/v1/ofo-public/

The data files are also accessible via S3 protocol (e.g. AWS S3 CLI, rclone, CyberDuck) at endpoint: https://js2.jetstream-cloud.org:8001/, bucket ofo-public.

The OFO web catalog contains web preview and direct download functionality for drone-derived data.

The OFO STAC catalog is an altnerative means of data access via the standard Spatio Temporal Asset Catalog specification, but it currently reflects an older processing run for a subset of missions. We plan to update it to reflect our current catalog by the end of 2026 and will prioritize this if/when requested by an external user.

Data types

• Geospatial orthomosaics are named <mission-id>_ortho-dsm-ptcloud.tif or <mission-id>_ortho-dsm-mesh.tif depending whether they were based on a digital surface model made from the point cloud (more detailed) or the mesh (smoother).

• Canopy height models are named <mission-id>_chm-ptcloud.tif or <mission-id>_chm-mesh.tif depending whether they were based on a digital surface model made from the point cloud (more detailed) or the mesh (smoother). All CHMs use the point cloud-derived DTM as their terrain model.

• The digital terrain model is made from the point cloud (only) and is named <mission-id>_dtm-ptcloud.tif.

• Each processed product should also have a color point cloud (3D), mesh model (3D) (filename contains “model”; both in an arbitrary project-specific coordinate system, or a global geospatial coordinate system), digital surface models (from mesh and point cloud), raw photo geospatial locations, camera poses, and processing report. The mapping of these data types to the files should be apparent based on the filenames.

• Each mission should also have detected treetops and crowns (the latter from each of two algorithms, Silva and Watershed). They are located in the itd-0001 folder beneath the mission’s photogrammetry_03 folder.

• You can stream the point cloud into the Entwine viewer by going to Entwine selecting “Add data” and entering the HTTP URL of the point cloud, such as: https://js2.jetstream-cloud.org:8001/ofo-public/drone/missions_01/000001/photogrammetry_03//full/000001_points-copc.laz

Important details and considerations

• The OFO drone data schema describing metadata attributes is in revision. Please reach out with any questions about metadata interpretation.

• Important usage restriction: We have designated a stratified subset of approximately 20% of drone missions to be excluded from training ML models that are intended for use by the broader forest mapping/ML community. For such applications, these missions are reserved for use as test data for model evaluation and intercomparison. The missions to be excluded from training are indicated by the value true in the metadata column withheld_from_training.

• Many OFO drone inventories consist of two missions: (1) high-altitude, nadir imaging and (2) low-altitude, oblique imaging. The latter is intended to complement the former and allow for enhanced multi-view tree detection and species classification from raw drone images. For applications that do not involve interpretation of raw drone images, the oblique missions should likely be excluded.

• Data license varies by mission. All use Creative Commons licenses requiring data contributor attribution in derivative products, but please be aware that some mission data licenses may include non-commercial and share-alike clauses that limit derivative works and their uses in certain contexts.

Important general considerations

• The mapping of drone missions to ground reference plots is not 1:1. Some ground plots are covered by multiple drone missions (or zero drone missions) and vice-versa.

• Ground reference plots are not necessarily spatially co-registered to the drone missions that include them. In most/all cases, they should be within ~10 m, but for most comparison analysis, it will likely be necessary to perform (manual or automated) co-registration to optimally align the ground data to the drone data.