Proportional estimates of cover are developed from global training data derived using high-resolution imagery. The training data and phenological metrics are used with a regression tree to derive percent cover globally. The model is then used to estimate areal proportions of life form, leaf type, and leaf longevity. The current V004 collection of the yearly MODIS Vegetation Continuous Fields (VCF) product contains only percent tree cover, with the other layers to follow in later releases.
The data layers in the VCF product are generated on an annual basis from monthly composites of 500-m Surface Reflectance data. Compositing is based on the second darkest albedo to remove clouds and cloud shadow. For more information refer to the MODIS VCF user manual.
The following library is intended to be used to accelerate the production development of MODIS VCF products. All dependencies have been housed inside OCI Docker containers that can be directly used with Docker, Singularity, Podman, among others.
Note: PIP installations do not include CUDA libraries for GPU support. Make sure NVIDIA libraries are installed locally in the system if not using conda/mamba.
module load singularity
singularity build --sandbox /lscratch/$USER/container/modis-vcf docker://nasanccs/modis-vcf:latestmodule load singularity
singularity build --sandbox /lscratch/$USER/container/modis-vcf-dev docker://nasanccs/modis-vcf:devBelow is the documentation to reproduce the validation of MODIS VCF training data using Very High Resolution imagery. The following directions are meant to assist in reproducing these steps for a future validation sceneario. The process is not fully automated given the need to select based on operator experience, the best scenes to work with. This documentation is work in progress.
- Mark L. Carroll, [email protected]
- Roger L Gill, [email protected]
- Savannah L Strong, [email protected]
- Jordan Alexis Caraballo-Vega, [email protected]
Please see our guide for contributing to modis_vcf.