Global map shows forest canopy height in shades of green from 0 to 70 meters (Source: NASA)
This product provides information on tree heights in tropical and boreal forests. The EO techniques use stereo data to produce a digital surface model (DSM) and digital terrain model (DTM), which are then combined to derive a normalised digital surface model (nDSM), which shows the height of all features above the Earth's surface. Land cover classification techniques are used to distinguish forest from other elevated land cover/use classes (e.g. shrubs or build-up) and the height of various landscape features (trees, shrubs, etc.) is calculated.
Tree height information can be used as a component to determine forest above-ground biomass. Furthermore, tree heights can be plugged into models that predict the spread and behavior of fires, as well as ecological models that help to understand the suitability of species to specific forests.
Known restrictions / limitations
Tree height or canopy height information is a standard forestry parameters. Input data sources are optical stereo data or a combination of optical (single image) and radar (stereo image). In tropical rain forest areas, frequent cloud cover can be an issue for deriving the forest cover information. If optical stereo data for generation of DSM, DTM and nDSM are used, the effects of clouds, cloud shadows as well as shadow areas caused by terrain, can lead to missing elevation information and this must be considered. LiDAR is the standard approach to derive this information as it delivers more accurate height information.
Lifecycle stage and demand
Pre-Licensing, Exploration & Development:
Geographic coverage and demand
Demand is global, in regions with dense forest cover.
Input data sources
Optical: VHR1, VHR2, HR1
Radar: VHR1, VHR2, HR1
Spatial resolution and coverage
Spatial resolution: 2 - 30 m pixel size
Minimum Mapping Unit (MMU)
n/a (the product is directly based on the input data; the smallest unit is one pixel)
Accuracy / constraints
The geometric accuracy is less than 1 pixel which in the case of tree cover density is on the order of 2-30 m.
Thematic accuracy: 80-90%
Spatial accuracy: The goal would be one pixel, but depends on reference data.
Accuracy assessment approach & quality control measures
Stratified random points sampling approach utilizing VHR reference or other geospatial in-situ data. Statistical confusion matrix with user’s and producer’s accuracy as well as kappa statistics for tree height.
Frequency / timeliness
Observation frequency: The frequency is constrained by satellite revisit and acquisition timeframes, but also processing requirements. Depending on the requirements of the customer, the most suitable satellite sensor has to be selected considering spatial / spectral resolution as well as revisit frequency. Typically, long-term changes are detected on a 3 to 5 year basis.
Timeliness of delivery: Depending on size of the mapped area, resolution, MMU.
VHR1, VHR2 and HR1 data must be commercially acquired.
Delivery / output format
Maria Lemper; Jan Militzer
# of Pages:
Internal – Project consortium and science partners
External – ESA
This page has no comments.