A spatially varying model for small area estimates of biomass density across the contiguous United States
Description
Data and code to reproduce the results in the like-named paper. Code is an R script that implements a Fay-Herriot model with spatial error and spatially varying regression coefficients to make estimates of above ground biomass density (AGBD) for 64,000 ha hexagons across the contiguous United States. Uses FIA direct estimates along with area predictors from the Global Ecosystem Dynamics Investigation (GEDI) and the NLCD Tree Canopy Cover (TCC) map. 'UShex_SFH_RSE.R' contains the code, and is well-commented, but those unfamiliar with INLA might want to refer to the free, online book https://becarioprecario.bitbucket.io/spde-gitbook/ in order to fully understand the code. The code, as is, fits the GEDI+TCC Spatial Fay-Herriot model, but predictors and spatial effects can be omitted/introduced according to the users desires. The later part of the script executes the 10-fold cross-validation study. 'GEDI_HEX.RData' contains the data necessary to fit the GEDI model. 'TCC_HEX.RData' contains the data necessary to fit the TCC model. Both are needed to fit the GEDI+TCC model. The variables inside are described in the comments of the R script. CONUSbiohex2020 is a folder with the shapebundle from Menlove & Healey, and is also available at https://daac.ornl.gov/CMS/guides/FIA_Forest_Biomass_Estimates.html#datadescraccess. It's not necessary for the analysis in the R script, but is useful for spatial plots, and for comparing to (or using) the post-stratification estimates.