Small area population estimates using random forest top-down disaggregation (Part 2): the popRF ‘R’ package

that the reference year of the population data is the same as the administrative boundary data and the spatial extent matches the enumerated population) - It is similarly important that the unique identifiers in the administrative boundary layer match the unique identifiers in the tabulated population data so that (i) every polygon in the administrative boundary layer has a corresponding record in. [...] Thus, the spatial extent of the covariates must be adjusted to the be exactly the same as the spatial extent of the administrative boundaries, and thence, the population data. [...] If, however, newer data is unavailable and the pre-prepared covariate is not available for the neighbouring countries, you need to spatially extrapolate the values in the pre-prepared covariate raster and then cookie-cut the correct area using the administrative unit boundaries to ensure that the spatial extents match exactly. [...] The required parameters are the paths to the above input files: • pop: path to the ‘Population Counts’ tabular file • mastergrid: path to the ‘Zonal Data’ raster file • cov: path to the covariate rasters (i.e. [...] An example model code is shown in Appendix A, but the main steps are listed here: - start R or Rstudio - load the popRF library - set the directory location of the inputs - create a list from the names and locations of the needed covariates - state the names and locations of the mandatory inputs - write the command line that executes the Random Forest application.