2.2. Pre-processing: Spatial data

As described in the section on Input data, a large number of spatial datasets are needed as input to run the model. Most of the data can be taken from global data products that are are available in the mapspamc database. In case more detailed national data sources (e.g. national land cover, accessibility or population maps) are available, the user can use those instead.

For this reason, we decided not to create specific functions to process the spatial data, which offer limited flexibility for user interaction. Instead we share the detailed scripts that we used to process all the spatial datasets, which can be found in the 02_2_pre_processing_spatial_data folder. To process raster data, we created the align_raster() function that can be used to clip country data from a global spatial layer using a country shapefile and at the same time reproject the map to the desired extent, resolution and coordinate reference system.

The code below illustrates the use of align_raster() to clip the Malawi data from the global WorldPop map used in 04_select_worldpop.R. Note that several data layers require additional or slightly different processing approach before they can be used. For example, the raw WorldPop dataset presents population density per grid cell at a resolution of 30 arc second. If we want to run the model at a resolution of 5 arc minutes, which has grid cells that are 100 times larger, the data needs to be summed when aggregating the grid cells. This is different from the default option in align_raster(), which uses a bilinear resampling approach when reprojecting the maps.

input <- file.path(param$db_path, glue("worldpop/ppp_{param$year}_1km_Aggregated.tif"))

if (param$res == "30sec") {
  temp_path <- file.path(param$model_path, glue("processed_data/maps/population/{param$res}"))
  dir.create(temp_path, showWarnings = FALSE, recursive = TRUE)

  output <- align_raster(input, grid, adm_map, method = "bilinear")
  plot(output)
}

if (param$res == "5min") {
  temp_path <- file.path(param$model_path, glue("processed_data/maps/population/{param$res}"))
  dir.create(temp_path, showWarnings = FALSE, recursive = TRUE)

  output <- align_raster(input, grid, adm_map, method = "average")

  # Multiply average population with 100
  output <- output * 100
  plot(output)
}

To process the spatial data, the user needs to run all scripts in the 02_2_pre_processing_spatial_data folder. Note that some of the scripts, in particular select_gaez.R, which processes more than 300 global maps, can take a long time to run!

The next step involves the creation of the synergy cropland map.