2.4. Pre-processing - irrigated area

Similar to the synergy cropland map, we combine two global irrigated area maps (GIA and GMIA) to create a synergy irrigated area product - for more information see the mapspamc database documentation. Scripts are provided in the 02_4_pre_processing_irrigated_area folder to process the two maps. The 03_create_synergy_irrigated_area_map.R script describes the process to create the synergy irrigated area components that are needed by mapspamc. The procedure is slightly different than that used to create a the synergy cropland map because we only have two sources of information, which also have different resolutions. Most problematic is the GMIA, which has a relatively coarse resolution of 5 arc minutes. This causes a problem when running the model at 30 arc seconds as we do not know which of 100 30 arc seconds grid cells that are located in a single 5 arc minutes grid cell, are irrigated. Another limitation is that even the GIA, the other global irrigated area product, with a resolution of 30 arc seconds, only indicates whether a grid cell is irrigated. It does not present information on the actual area of cropland that is irrigated.

In order to combine the GMIA and GIA data, we processed both maps so that they present the share of irrigated area within a grid cell (select_*.R scripts). To create the synergy cropland map, we use the following procedure: (1) We stack the GMIA and GIA information and remove grid cells with very small shares (<0.01). These small values are sometimes created when reprojecting the maps to a finer or coarser resolution and can safely be removed; (2) We take the take the maximum of the GMIA and GIA irrigated area share; (3) We add a rank from 1 (highest share) to 10 (lowest share) using equal intervals of 0.1 irrigated area share; (4) We rank the cells from 1 to 10 and; (5) We calculate the irrigated area in ha.

ir_df <- as.data.frame(c(grid, grid_size, gmia, gia), xy = TRUE) %>%
  filter(!is.na(gridID)) %>%
  dplyr::select(-x, -y) %>%
  mutate(
    gia = ifelse(gia < 0.01, 0, gia),
    gmia = ifelse(gmia < 0.01, 0, gmia)
  )

ir_df <- ir_df %>%
  dplyr::mutate(
    ir_max = pmax(gmia, gia, na.rm = T),
    ir_rank = cut(ir_max,
      labels = c(1:10), breaks = seq(0, 1, 0.1),
      include.lowest = T
    ),
    ir_rank = dense_rank(desc(ir_rank)),
    ir_max = ir_max * grid_size
  ) %>%
  filter(!is.na(ir_rank), ir_max > 0) %>%
  dplyr::select(-gmia, -gia, -grid_size)

By ranking on irrigated area share, GIA (share is 1) is always preferred over GMIA when a resolution of 30 arc seconds is selected by the user. In case a resolution of 5 arc minutes is used, GMIA and GIA grid cells with a large share of irrigated area receive a higher rank, which is also desirable. In the remainder of the script, the maps for maximum irrigated area and ranking are created.