robitalec / move-book Goto Github PK

Movebank dataset summarizer

This project scrapes and summarizes data sets available on Movebank. There is a huge diversity of data sources with varying license terms, public access/not, taxa. This is an attempt to better understand how accessible relevant data are for my own project. I'm hoping to wrap this up as a more general-use template, so if that's something you would be interested in - please open an issue and we can discuss. I'd rather do that with an idea of what others are looking for. In the meantime, this can be used and adapted for your own needs.

We use a number of packages for downloading (rmoveapi, move), processing (data.table, anytime) and visualizing (ggplot2, leaflet) the data. All steps are wrapped up in a targets workflow, and package versions tracked with renv. The result is a bookdown doc with each study on its own page. There's a minimum working example of combining targets and bookdown available here: robitalec/targets-parameterized-bookdown. Thank you to the developers of all of these great packages.

1. Register for a Movebank account
2. Run the setup script to install packages with renv, save your credentials with keyring and download study data
3. Edit the credentials and path sections in the targets file (_targets.R)
4. Run targets::tar_make()

I've decided to remove the study data download step from the targets workflow because during development it was a huge step I didn't want to rerun. I figure folks will do this step once, so it's not worth hanging up the workflow. Since the paths are tracked however, targets will only rerun what is needed as data are added or removed.

If you are hoping to do something similar, I have two main lessons learned:

• Reduce the data to only what you need as quickly as possible, without going far into analysis or processing. This is directly related to...
• The data is not homogeneous. Data types do not appear to be strictly enforced, there are many duplicated study names, and there are NAs or errors throughout. This makes it hard to apply the same functions, or combine different datasets programmatically.

So reduce down to only the data you need before doing anything too specific.

Some things to watch out for:

• duplicated study names (see here)
• errors in dates (eg. typos, impossible study period date/end ranges)
• taxons provided are often a list of different taxonomies or taxonomic ranks
• pay attention to the Movebank API doc

This project has been released under a GPL-3 license and with a Zenodo DOI:

Alec L. Robitaille. robitalec/move-book. Zenodo.