D Case Study - Alternate PM_2.5 surfaces

The PM2.5 exposure surfaces used in the provided pipeline comes from Surface PM_2.5 V4.GL.02 dataset developed by van Donkelaar, A. et al. (2016). Alternative models of PM_2.5 differing in method, temporal or spatial coverage and/or resolution which are more suited to your particular study may be used as the input exposure rasters.

To use different PM_2.5 surfaces, replace the call to function import_globalgwr_pm25_2010_2015 in _targets.R with targets importing and tidying the new raster data.

A satellite-derived land use regression (SatLUR) modelled PM2.5 surface produced by Luke Knibbs is available on request.

---

The preparation of van Donkelaar’s PM_2.5 surfaces is performed in targets infile_globalgwr_pm25_2010_2015_files, infile_globalgwr_pm25_2010_2015 and tidy_env_exposure_pm25. (Only the latter two are present if downloading the data.) The first two targets are automatically generated by the tarchetypes::tar_files_input then tidied and gathered into a raster brick in tidy_env_exposure_pm25. This is performed in the following few lines:

inputs <- list(
  ...,

  # exposure rasters
  exposure = import_globalgwr_pm25_2010_2015(years, 
                                             download = download_data, 
                                             datadir_envgen = file.path(dir_cardat, dir_envgen)), 
  ...
)

Substitute in a tar_files_input pointing to the new input exposure raster files. Read and tidy the data in a tar_target, producing a RasterBrick (or stack) with named layers by year. Your code should look similar to the following.

inputs <- list(
  ...,

  # exposure rasters
  exposure = list(
    tar_files_input(
      infile_pm25,
      file.path(
        sprintf("YOUR_DATA_DIR/SatPM25_2000_2015/data_derived_rasters/satlur_pm25_ug_m3_%s.tif",
                years)
      )
    ),
    tar_target(
      tidy_env_exposure_pm25,
      {
        b <- brick(stack(infile_pm25))
        # rename the layers of the RasterBrick with the year
        names(b) <- years
        return(b)
      }
    )
  ),
  ...
)

Add additional processing steps in tidy_env_exposure_pm25 as needed to tidy the raster input.