In dbca-wa/etlTurtleNesting: Data ETL of Turtle Nesting Census Data from ODK Central to WAStD

library(wastdr)

if (is.null(params$w2_data)){
  wastdr::wastdr_msg_info("Using locally saved database snapshot")
  w2_data <- readRDS(here::here("inst/w2.rds"))
  # w2_data <- drake::loadd("w2_data")
} else {
  wastdr::wastdr_msg_info("Using data via vignette parameters")
  w2_data <- params$w2_data
}

QC for WAMTRAM geolocation

The following maps show the hand-typed coordinates of WAMTRAM encounters for one location code (e.g. Thevenard Island or Delambre Island), as well as encounters for one place code (e.g., THV tagging area, Delambre general).

For performance, place markers are grouped, expanding on zoom.

Error-free data would be contained entirely on the respective nesting beaches.

Thevenard Island

Location TH

map_wamtram(w2_data, location = "TH")

Thevenard Island East End (THEE)

map_wamtram(w2_data, place = "THEE")
# map_wamtram(w2_data, place = "THNB")

Delambre Island

Location DA (Dampier)

WAMTRAM encounters linked to site "DADI" (Delambre Island) are linked to the following location codes: r w2_data$enc %>% dplyr::filter(place_code == "DADI") %>% magrittr::extract2("location_code") %>% unique(). All but "DA" are wrong.

map_wamtram(w2_data, location = "DA")

Delambre Island (DADI)

The most impressive migration is shown by the Flatback turtle off Namibia seen in 2014. Others have made it to Micronesia - at least per typo.

map_wamtram(w2_data, place = "DADI")

Barrow Island

map_wamtram(w2_data, location = "BW")

Dirk Hartog

Location DH

Dirk Hartog as a rookery spans most of WA, up to the Top End, and down to the South coast.

map_wamtram(w2_data, location="DH")

DHI Turtle Bay (DHTB)

Data from the site DHTB is actually pretty nicely georeferenced. The above map shows records that might be included below but currently are missing due to an incorrect place code.

map_wamtram(w2_data, place="DHTB")

Conclusion

The sheer number of misplaced hand-typed coordinates with errors not only in coordinates, but also in locatino and place codes, means that significant human QA effort would be required to fix all of these coordinates. The star-pattern shows that misplaced coordinates mostly have a typo in one digit of one coordinate. This also means that coordinates within a site's boundaries can be off by 10 or 100 metres, rendering the association to the underlying habitat unreliable.

Mapping of WAMTRAM to WAStD sites

The following map shows WAStD sites (blue polygons) and WAMTRAM sites (points). WAMTRAM sites are either linked to WAStD sites (green markers), or not yet linked (red markers).

Once all WAMTRAM sites are linked to exactly one WAStD site, the map below will show only green place markers, and we can approximate coordinates for WAMTRAM encounters from the referenced site. In detail, we will use the coordinates of the centroid of the WAStD site as the coordinates of the WAMTRAM encounter.

This will

• lose both precision and typo errors of all hand-typed coordinates,
• assume the place code is correct (above maps show that's not always the case),
• allow the import of all WAMTRAM encounters which are at least georeferenced by place name.

areas_sf <- wastd_GET("area") %>% 
  magrittr::extract2("data") %>%
  geojsonio::as.json() %>% 
  geojsonsf::geojson_sf()

areas <- areas_sf %>%
 dplyr::filter(area_type == "Locality") %>%
 dplyr::transmute(
   area_id = pk, 
   area_name = name, 
   w2_location_code = w2_location_code
 )

sites <- areas_sf %>%
 dplyr::filter(area_type == "Site") %>%
 dplyr::transmute(
   site_id = pk, 
   site_name = name, 
   w2_place_code = w2_place_code
 ) %>% 
 sf::st_join(areas)

wastdr::map_wastd_wamtram_sites(areas, sites, w2_data$sites)

Stats of WAMTRAM encounters:

• r w2_data$enc %>% dplyr::filter(is.na(place_code)) %>% nrow() encounters with hand typed coordinates lack a place code,
• r w2_data$enc %>% dplyr::filter(is.na(place_code), is.na(location_code)) %>% nrow() encounters with hand typed coordinates lack both a place and a location code,
• r w2_data$enc_qa %>% dplyr::filter(is.na(place_code)) %>% nrow() encounters without hand typed coordinates also lack a place code
• r w2_data$enc %>% dplyr::filter(is.na(place_code), is.na(location_code)) %>% nrow() encounters without hand typed coordinates lack both a place and a location code,
• r w2_data$enc_qa %>% dplyr::filter(is.na(place_code), is.na(location_code), is.na(longitude_degrees)) %>% nrow() encounters without hand typed coordinates lack both a place, a location code, and have no DMS coordinates (ideally 0).

sites_nocoord <- w2_data$sites %>%
  dplyr::filter(is.na(site_longitude)) 

Furthermore, there are r sites_nocoord %>% nrow() WAMTRAM sites without coordinates. All of these need to be linked to new or existing WAStD sites. Some may be known from their place names.

sites_nocoord %>% reactable::reactable()

Conclusion

All errors above need to be rectified before WAMTRAM tagging data will be complete, consistent, and correct in WAStD.

While QA progresses, we can already report on WAMTRAM tagging data, but will see changes to historic data until QA is completed.

dbca-wa/etlTurtleNesting documentation built on Nov. 18, 2022, 8:03 a.m.