Canopy height model and NAIP imagery pairs across CONUS

Area of Science:

• Geospatial science
• Remote sensing
• Ecology

Background:

• Canopy height models (CHM) offer crucial environmental vertical structure data for ecological and geospatial applications.
• Existing CHMs often lack spatiotemporal consistency, limiting their use and requiring scaling.
• Scaling CHMs is challenging due to insufficient spatially diverse data.

Purpose of the Study:

• To generate a comprehensive dataset of CHM chips to facilitate the development of models for large-scale CHM production.
• To overcome data limitations hindering the spatiotemporal scaling of CHMs.

Main Methods:

• Utilized United States Geological Survey 3D Elevation Program lidar data to produce 22,796,764 one-meter resolution CHM chips.
• Stratified CHM chips across dominant land cover types in the conterminous United States.
• Paired each CHM chip with a time-aligned aerial image from the United States Department of Agriculture National Agriculture Imagery Program.

Main Results:

• Generated a massive dataset of 22,796,764 high-resolution CHM chips.
• Created paired CHM and aerial imagery data covering diverse land covers across the U.S.
• Established a foundation for training models to produce large-scale CHMs.

Conclusions:

• The created dataset addresses the need for spatially diverse data to improve CHM spatiotemporal consistency.
• This resource is pivotal for advancing research and applications reliant on accurate and scalable CHMs.
• Enables the development of advanced machine learning models for automated, large-scale CHM generation.