The GEDI Mission Team has corrected an error in the GEDI Level 4A Footprint Level Aboveground Biomass Density (AGBD) Version 3 product that affects AGBD estimates for South American (SA) evergreen broadleaf trees (EBT) and deciduous broadleaf trees (DBT) strata. A corrected version of the GEDI Level 4A Footprint Level Aboveground Biomass Density data product will be released.
The error was caused by a bug in the model fit code, resulting in incorrect model selection and predictions for shots using the EBT_SA model. The EBT_SA model is used to predict AGBD in two South American strata: EBT and DBT forests. These predictions should not be used. Shots in other strata are not affected. This bug did not impact Version 2 model selection and predictions.
The rest of this data alert describes how users can generate new estimates prior to release of the corrected data product.
What is Affected?
Only the model coefficients for the EBT_SA model have changed. The predictors (the relative-height metrics) used by the model are unchanged. Because the predictors are unchanged, users wishing to generate corrected estimates before the updated data product is released can use the predictors already stored in each Level 4A granule for each shot rather than re-deriving the predictors from the GEDI Level 2A product.
Recommended Action
The GEDI Mission Team is preparing a corrected version of the GEDI L4A data product that will incorporate the corrected model coefficients for EBT_SA and DBT_SA prediction strata.
Users wishing to generate corrected estimates before the new data product is released can use one of two ready-to-run notebooks—one notebook in Python and one notebook in R—to predict AGBD using the corrected EBT_SA model. These notebooks illustrate how to use data already stored in each L4A granule and do not require independent access to GEDI L2A data.
Please direct any questions or concerns regarding this issue to uso@daac.ornl.gov.