O_Vlieland: add behaviour data based on acceleration data

[peterdesmet]

From @henkjanvdkolk

O_Vlieland wil ik graag ook geclassificeerde gedragingen bij de dataset voegen, maar dit kan waarschijnlijk alleen op Zenodo en niet op Movebank? Dat wordt dan een file met kolommen BirdID(of tracker ID) + date_time_stamp + geclassificeerd gedrag

[peterdesmet]

It looks like this type of data can be uploaded to Movebank, as two fields:

• behavioural classification, examples roosting, foraging, running
• behaviour according to, example Behavior was derived using acceleration data and an automated algorithm calibrated using field observations.

The data itself should then be linked to the position record.

@henkjanvdkolk

1. Do you have that data at the level of position records?
2. Where is that data currently stored? In UvA-BiTS?

[peterdesmet]

Update:

• Data is not stored on UvA-BiTS
• Data is available for every start-timestamp in acceleration data (although some years or not fully covered).
• Acceleration data were collected for more start-timestamps (8,021,727 records) then positions were recorded (4,829,950)

After talking with @sarahcd:

• Don't associate with GPS, since that would mean losing a lot of data
• Don't associated with ACC, since that would inflate records (there are 10 rows for each start-timestamp) and make it harder for users to download that data
• Do upload data as Accessory Measurements. Advantages: tidy data, clear for user there is this separate data source, easy to download that data specifically, we can store it as such on Zenodo, users can link GPS + behaviour based on timestamp themselves.

[peterdesmet]

Code that was used to clean up the behaviour data, ending up with 6977784 records.

library(tidyverse)
library(here)
library(lubridate)

# Load behaviour data
beh <- read_csv("data/processed/O_VLIELAND/bird_ACC_2_v2.csv")
#> Rows: 7077471 Columns: 7
#> ── Column specification ────────────────────────────────────────────────────────
#> Delimiter: ","
#> chr  (1): class_name
#> dbl  (5): device_info_serial, longitude, latitude, altitude, gpsspeed
#> dttm (1): date_time
#> 
#> ℹ Use `spec()` to retrieve the full column specification for this data.
#> ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
nrow(beh)
#> [1] 7077471

# Remove duplicate rows
beh <- beh %>% distinct(.keep_all = TRUE)
nrow(beh)
#> [1] 7038045

# Create standardized values and column names
beh <-
  beh %>%
  mutate(
    `tag-id` = device_info_serial,
    timestamp = paste0(str_replace(date_time, " ", "T"), "Z"),
    timestamp_dttm = ymd_hms(timestamp),
    `behaviour-classification` = tolower(class_name)
  ) %>%
  select(`tag-id`, timestamp, timestamp_dttm, `behaviour-classification`)

# Load reference data and set future deploy-off-timestamp if NULL
ref <- read_csv("data/processed/O_VLIELAND/movebank_ref_data.csv")
#> Rows: 103 Columns: 25
#> ── Column specification ────────────────────────────────────────────────────────
#> Delimiter: ","
#> chr  (13): animal-ring-id, animal-sex, animal-taxon, animal-life-stage, atta...
#> dbl   (8): animal-id, animal-mass, deploy-on-latitude, deploy-on-longitude, ...
#> lgl   (2): animal-comments, animal-nickname
#> dttm  (2): deploy-off-timestamp, deploy-on-timestamp
#> 
#> ℹ Use `spec()` to retrieve the full column specification for this data.
#> ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
ref <- ref %>%
  mutate(`deploy-off-timestamp` = if_else(
    is.na(`deploy-off-timestamp`),
    as.POSIXct("2100-01-01T00:00:00Z", tz = "UTC"),
    `deploy-off-timestamp`
))

# Join behaviour data with reference data
# Is initially going to increase number of records for tags associated with 2
# animals, but those will disappear when selecting on timestamp within session
beh_ref <- beh %>% left_join(ref, by = "tag-id")
beh_ref <- beh_ref %>% filter(
  (timestamp_dttm >= `deploy-on-timestamp`) &
  (timestamp_dttm <= `deploy-off-timestamp`)
)
nrow(beh_ref)
#> [1] 6977784

# Export data
beh_ref <- beh_ref %>% select(`tag-id`, `animal-id`, `timestamp`, `behaviour-classification`)
write_csv(beh_ref, here::here("data", "processed", "O_VLIELAND", "movebank_beh.csv"), na = "")

[peterdesmet]

Behaviour data are now uploaded as Accessory Measurements: https://www.movebank.org/cms/webapp?gwt_fragment=page=studies,path=study1605802367