BioHack_iNat

Biohackaton 2021 - Team investigating recorder behaviour on iNaturalist

• BioHack_iNat

Getting metrics for the mean Spanish plant collector

Building a list of observers

The goal is the create metrics for a list of 'mean' observers. We created a gbif export for research grade inaturalist observations, in spain, since 2016. Then filtered on observers with atleast 100 observations since then.

The export is here: GBIF.org (09 November 2021) GBIF Occurrence Download https://doi.org/10.15468/dl.svwnrp

This results in 461 observers meeting this criteria.

Getting a data dump from inat

Make a directory called data_inat in the main repo directory and use aws command line tool (https://aws.amazon.com/cli/) to download data:

mkdir data_inat
cd data_inat
aws s3 cp s3://inaturalist-open-data/observations.csv.gz observations.csv.gz --no-sign-request
aws s3 cp s3://inaturalist-open-data/taxa.csv.gz taxa.csv.gz --no-sign-request

The data_inat directory is git ignored so will not be committed/pushed. This data is currently not used in the workflow.

Getting all inat observations for these users

This part of the process is run in different code chunks in https://github.com/AugustT/BioHack_iNat/tree/main/2_Generating_iNat_RM_Baseline

From the 461 observers from GBIF - we don't actually have their usernames, we have a recordedBy column which is a mix of names and usernames. Therefore we need to resolve these to iNaturalist usernames. We can do this by using the endpoint: https://api.inaturalist.org/v1/docs/#!/Users/get_users_autocomplete

If multiple users are found by this method we currently just drop those users. This leaves us with 319 users.

Next we make a query to the iNaturalist API https://api.inaturalist.org/v1/docs/#!/Observations/get_observations but it is wrapped in the get_inat_obs_user() function in R package rinat https://cran.r-project.org/web/packages/rinat/index.html. However in order to filter by location and species I have edited the function to access extra bit of the query, which for plants in Spain is "&taxon_id=47126&place_id=6774".

This query is run for all users with a maximum number of records as 5000 - some observers do have over 5000 records so this is just an arbitrary limit whilst we're hacking. This results is then saved as a dataframe as a .rds object in https://github.com/AugustT/BioHack_iNat/tree/main/2_Generating_iNat_RM_Baseline/data/user_obvs

These are then all loaded in and combined into one dataframe using

all_observations_raw <- paste0("2_Generating_iNat_RM_Baseline/data/user_obvs/",list.files(path = "2_Generating_iNat_RM_Baseline/data/user_obvs", pattern = ".rds")) %>%
  map(readRDS) %>% 
  bind_rows()

Preparing data for recordMetrics

Appropriate columns are selected:

• id - the observation id
• recorder - the recorder username
• species - the species scientific name
• date - the date in YYYY-MM-DD format and converted to date using as.Date()
• long - decimal longitude
• lat- decimal latitude
• quality_grade - not essential for recorder metrics but possible useful

We also derive these columns

• species_level - is the record identified to species level? (determined in a hacky was by checking if there's a space in the species column
• country - country determined by using function coords2country() function defined in coords2country.R, then used to remove any records that are not from the target country
• km_sq - the km square (stored as character) derived by converting points to spanish national grid and rounding to nearest 1000 metres

This file is then saved to 2_Generating_iNat_RM_Baseline/data/observations/baseline_observations_spain_plants.rds and is ready for plugging into recorderMetrics package.

Calculating metrics