In sustainablefarms/sflddata: Data Extraction Methods for the Linking Data Project

Need to check:

• names in Garnett's data match the names in our data
• the removal set given by Garnett's data is consistent with the removal set given by Ikin's data

library(dplyr); library(tidyselect); library(tidyr);
knitr::opts_knit$set(root.dir = rprojroot::find_root(rprojroot::has_file("DESCRIPTION")))
devtools::load_all(rprojroot::find_root(rprojroot::has_file("DESCRIPTION")))

traits <- read.csv("./private/data/raw/Australian_Bird_Data_Version_1.csv")
traits_ikin <- as.data.frame(
  readxl::read_excel(
    "./private/data/raw/Ikin_SWS_Bird_Traits_updatedApril2017.xlsx",
    sheet = "Ikin_SWS_Bird_Traits"
  ))
birds_raw <- readRDS("./private/data/raw/birds_long.rds")
birds_raw$CommonName <- gsub("Horsfield\x92s Bushlark", "Horsfield's Bushlark", birds_raw$CommonName) #weird encoding of a bird name
birds_raw <- birds_raw %>% dplyr::filter(CommonName != "Corella sp.") #remove the single observation of Corella sp. as this could be at least two different common names.

The names in Garnett's data match the names in our data

Unique names for rows

Column's in Garnett's data that make for good names are unique to each taxon (each row), and similar common names or scientific names in our data.

anyDuplicated(traits$X3_Taxon_common_name_2)
anyDuplicated(traits$X5_Species_name_2)
anyDuplicated(traits[, c("X5_Species_name_2", "X6_Subspecies_name_2")])
anyDuplicated(traits$X10_Family_scientific_name_2)
anyDuplicated(traits$X17_Species_2)
anyDuplicated(traits[, c("X17_Species_2", "X18_Subspecies_2")])

In fact only the index and X3_Taxon_common_name_2 are unique to each row:

which(apply(traits, 2, anyDuplicated) == 0)

X3_Taxon_common_name_2 is also quite similar to our common names (see below).

Mapping to our CommonName

Generalised Levenshtein distances between names

namedists <- adist(unique(birds_raw$CommonName), traits$X3_Taxon_common_name_2)
rownames(namedists) <- unique(birds_raw$CommonName)
colnames(namedists) <- traits$X3_Taxon_common_name_2

Summarising the row minimums shows that nearly every common name we have in our data is within (approximately) 1 substitution from a name in the Australia Birds data set

summary(factor(apply(namedists, 1, min, na.rm = TRUE)))

Create a mapping by get the closest taxon names for each of our common names

NameMap <- apply(namedists, 1, function(v){
  ind <- which.min(v)
  return(names(v)[[ind]])
})

names of NameMap are the common name that we use in our data values of NameMap are the common names in the Garnett et al data set

These species have a difference of 1 r NameMap[apply(namedists, 1, min) == 1].

These names have a higher difference r NameMap[apply(namedists, 1, min) > 1].

Transform our data:

birds_renamed <- birds_raw %>%
  mutate(CommonName = NameMap[CommonName])

A quick note on Corella Sp

Corella sp has been removed because it could be multiple species

grep("Corella", unique(birds_renamed$CommonName), ignore.case = TRUE, value = TRUE)
grep("Corella", traits$X3_Taxon_common_name_2, ignore.case = TRUE, value = TRUE)

Birds to Exclude (and comparison to using Ikin's)

Exclude birds taxonimically: raptors and obligate water birds. We will also exclude the Australian Reed-Warbler, which is a passerine, because it requires wetlands to persist.

excluded_orders <- c(
  "Caprimulgiformes", # Frogmouths
  "Accipitriformes", # Eagles, Kites, Goshawks and Osprey
  "Strigiformes", # Masked Owls
  "Falconiformes", # Falcons
  "Podicepiformes", # Greebes
  "Gruiformes", # Crakes, Rails and Swamphens
  "Pelecaniformes", # Herons, Egrets, Pelicans etc
  "Anseriformes" # Ducks, Geese
)

birds_to_exclude <- as_tibble(traits) %>%
  dplyr::filter((X11_Order_2 %in% excluded_orders) | (X3_Taxon_common_name_2 == "Australian Reed-Warbler")) %>%
  dplyr::distinct()

# Birds to Include if Observed
birdskeep <- traits[!(traits$X1_Taxon_sort_1 %in% birds_to_exclude$X1_Taxon_sort_1), ]

To compare to Ikin need to create a map between species (taxon) names

namedists_ikin <- adist(traits_ikin$species, traits$X3_Taxon_common_name_2)
rownames(namedists_ikin) <- unique(traits_ikin$species)
colnames(namedists_ikin) <- traits$X3_Taxon_common_name_2
NameMap_ikin <- apply(namedists_ikin, 1, function(v){
  ind <- which.min(v)
  return(names(v)[[ind]])
})
# names of NameMap_ikin are the common name that Ikin's traits use
# values of NameMap_ikin are the common names in the Garnett et al data set
# The following have a difference of 1
NameMap_ikin[apply(namedists_ikin, 1, min) > 0]

Transform Ikin:

traits_ikin$species <- NameMap_ikin[traits_ikin$species]

Compare these to Ikin

species_to_remove_ikin <- traits_ikin %>%
  filter((diet == "Vertebrates") |
           (substrate == "Water") |
           (species == "Australian Reed-Warbler")) %>%
  dplyr::select(species) %>%
  arrange(species) %>%
  unlist() %>%
  as.vector()
species_to_keep_ikin <- setdiff(traits_ikin$species, species_to_remove_ikin)

Compare species removed

nearly all birds we want to remove based on Ikin's traits are in the birds to exclude list based on Garnett's data

summary((species_to_remove_ikin %in% birds_to_exclude$X3_Taxon_common_name_2))
species_to_remove_ikin[!(species_to_remove_ikin %in% birds_to_exclude$X3_Taxon_common_name_2)]

Compare species not removed using Ikin

How about species suggested to keep using Ikin's traits, but excluded by the new method above?

summary(species_to_keep_ikin %in% birds_to_exclude$X3_Taxon_common_name_2)
species_to_keep_ikin[(species_to_keep_ikin %in% birds_to_exclude$X3_Taxon_common_name_2)]
birds_renamed %>%
  dplyr::filter(CommonName %in% species_to_keep_ikin[(species_to_keep_ikin %in% birds_to_exclude$X3_Taxon_common_name_2)]) %>%
  group_by(CommonName) %>%
  summarise(Nobsatvisits = n()) %>%
  arrange(-Nobsatvisits)

11 species kept using Ikin's traits are removed using the new method. They look like ducks and water birds. Only 6 of these are observed enough to plausibly be placed in our model.

The following compare these 11 species on some functional traits: they all feed on inland waters. Some eat vertebrates.

library(ggplot2)
birds_to_exclude %>%
  dplyr::filter(X3_Taxon_common_name_2 %in% species_to_keep_ikin) %>%
  dplyr::select(X3_Taxon_common_name_2, X169_Food_Terrestrial_vertebrates_10,
                  X157_Breeding_habitat_inland_wetland_9,
                  matches("_Feeding_habitat_Inland_waters_")) %>%
  pivot_longer(-X3_Taxon_common_name_2) %>%
  dplyr::filter(value == 1) %>%
  ggplot() +
  geom_point(aes(y = X3_Taxon_common_name_2, x = name, col = name))

7_2_xx Modelled Species with Ikin Traits

modeldata <- readRDS("~/dlinkingdatacode/private/data/clean/7_2_10_input_data.rds")
modelledspecies <- modeldata$species
length(setdiff(modelledspecies, traits_ikin$species))

traits_ikin[traits_ikin$species %in% modelledspecies, ]

All species modelled have traits in the traits_ikin data! That is very convenient for functional diversity!

Compare to DBO Traits

source('../../../ComputerRelated/AccessLindenmayerSQL.R', echo=FALSE)

dbo_traits <- dbGetQuery(con, 
"SELECT sp.CommonName, bs.*
FROM dbo.BirdSpecies AS bs
INNER JOIN dbo.Species AS sp
ON bs.SpeciesId=sp.SpeciesId")
DBI::dbDisconnect(con)

all(modelledspecies %in% dbo_traits$CommonName)
dbo_traits %>%
  dplyr::filter(CommonName %in% modelledspecies) %>%
  dplyr::select(c(CommonName, ends_with("Score"))) %>%
  filter_all(any_vars(is.na(.)))

There are five birds that don't have any trait scores information: White-throated Gerygone
Yellow-faced Honeyeater
White-plumed Honeyeater
White-plumed Honeyeater
Diamond Firetail

dbo_traits %>%
  dplyr::filter(CommonName %in% c(
    "White-throated Gerygone",          
    "Yellow-faced Honeyeater",          
    "White-plumed Honeyeater",          
    "White-plumed Honeyeater",          
    "Diamond Firetail"
  )) %>%
  dplyr::filter(PresentInRegion == "1") %>%
  dplyr::select(c(CommonName, ends_with("Score")))

Some CommonNames occur twice or three times! Many of species present in region versions have scores. Of the 5 common names, only Yellow-faced Honeyeater has no scores.

sustainablefarms/sflddata documentation built on April 19, 2022, 11:19 a.m.