@angehung5
As discussed, this would be an effort to improve the granularity, and/or consistency between the input vegetation type dataset and the input GEDI canopy height and PAVD profile datasets.
Currently, as shown in previous issues/PRs, the GEDI height of maximum PAVD value does not match well with Massman prescribed heights in forest regions. This is likely impacted by surrounding low-lying vegetation in the original 1-km GEDI dataset, which when regridded (using dominant approach) to coarser resolution (e.g., GFS 13 km) results in a different profile shape compared to what a true forest point would look like at the sites compared in the canopy-wind paper (i.e., sites forced to be a forest vegtype). While a single point it is clear the issue, and to thus not use a regridded (dominant approach) GEDI 1-to-13 km PAVD product for the forest site, when applying to the global canopy-app it is clear such inconsistencies will arise.
Ultimately, we should bring in Barry's methods that uses the VIIRS 20 category 30 arcsec product to obtain a global gridded fractional vegtype dataset (from native high resolution, to a lower target grid resolution, e.g., 13 km), and adopt canopy-app to use it. Here we would need to change canopy-app to calculate a "sub-grid" fractional profile for each vegtype contained within the 13 km GFS grid, and then normalize profiles to the total number of vegtypes in the grid box. This would better represent the true heterogeneity in the vegtypes in each grid box and likely better match up with the originally higher resolution GEDI 1 km products.
@angehung5 As discussed, this would be an effort to improve the granularity, and/or consistency between the input vegetation type dataset and the input GEDI canopy height and PAVD profile datasets.
Currently, as shown in previous issues/PRs, the GEDI height of maximum PAVD value does not match well with Massman prescribed heights in forest regions. This is likely impacted by surrounding low-lying vegetation in the original 1-km GEDI dataset, which when regridded (using dominant approach) to coarser resolution (e.g., GFS 13 km) results in a different profile shape compared to what a true forest point would look like at the sites compared in the canopy-wind paper (i.e., sites forced to be a forest vegtype). While a single point it is clear the issue, and to thus not use a regridded (dominant approach) GEDI 1-to-13 km PAVD product for the forest site, when applying to the global canopy-app it is clear such inconsistencies will arise.
Ultimately, we should bring in Barry's methods that uses the VIIRS 20 category 30 arcsec product to obtain a global gridded fractional vegtype dataset (from native high resolution, to a lower target grid resolution, e.g., 13 km), and adopt canopy-app to use it. Here we would need to change canopy-app to calculate a "sub-grid" fractional profile for each vegtype contained within the 13 km GFS grid, and then normalize profiles to the total number of vegtypes in the grid box. This would better represent the true heterogeneity in the vegtypes in each grid box and likely better match up with the originally higher resolution GEDI 1 km products.