Download raw 2020 Island Areas Census DHC block group tables

Usage

download_bg_islandareas_raw(
  tables = islandareas_tables_for_bg_acsdata(),
  areas = c("AS", "GU", "MP", "VI"),
  key = Sys.getenv("CENSUS_API_KEY", unset = ""),
  download_fun = census_api_json_table,
  metadata_fun = census_api_group_variables
)

Arguments

tables: 2020 Island Areas Census DHC table groups to download, such as "P1".
areas: Island Area postal abbreviations to include.
key: optional Census API key. Defaults to CENSUS_API_KEY.
download_fun: function used to download one API URL. Defaults to census_api_json_table().

Value

list with raw blockgroup Island Areas Census DHC table lists plus metadata.

Details

This is a draft raw-data stage for AS/GU/MP/VI. These areas are not included in ACS, so the closest Census source is the 2020 Island Areas Census Detailed Housing Characteristics data. The returned object mirrors the bg_acs_raw table-list shape closely enough for later formula-mapping work. The 2020 Island Areas Census DHC source is not methodologically identical to ACS 5-year data. Some detailed age-by-sex fields are also not populated for most block groups, so pctfemale, pctunder5, pctunder18, and pctover64 are stored as NA when the source values are unavailable for Island Areas rows.

The annual EJScreen-compatible pipeline uses the archived EPA EJScreen ACS2022 reference as the default source for AS/GU/MP/VI row IDs, area fields, and available environmental fields. The transformed DHC demographics can be saved as bg_islandareas_demographics for review, but are not used in bg_acsdata unless use_islandareas_demographics = TRUE. The default path appends Island Areas rows with no DHC-derived demographic values so it remains compatible with legacy EPA/EJScreen Island Areas rows. That supports blockgroup dataset, EJSCREEN export, and map-data visibility for AS/GU/MP/VI. Island Area blocks are not added to the block helper datasets in the v3 release path, so point-buffer/radius analyses there should return no-data results rather than block-weighted estimates.