This repo is a collection of scripts and notebooks to download, merge and analyze data from GlobalFishingWatch.
We also process, split and build a classification model to predict wether a vessel encounter is authorized, partially authorized or not authorized.
Create an account and download all the Carriers v* zips from GlobalFishingWatch into a folder called carriers_data in the root of the repository.
Run the following command to unzip the files we are interested in:
python3 extract_zips.py
It will create a new folder called extracted that will contain the extracted files, and a file final.csv containing the merge of all the files.
The final.csv file is the one we will use to train our model.
Figure 1 is a plot showing 100k vessel encounters. As stated above, encounters are Authorizations are either known, partially known or unknown.
Encounters are locations where two vessels, a fishing and a carrier vessel, were within 500 meters for at least two hours and traveling at a median speed or less than 2 knots, while at least 30km away from a coastal anchorage.
