Original monitor data: history, states and confidence

What this dataset adds

The archive makes it possible to study the observatory itself: how often it published uncertainty, which operational categories it used and what confidence it assigned. That helps reveal whether the system tends to preserve old states, whether false positives are corrected and whether risk is kept separate from an effective closure.

Entries do not represent equal time intervals. A row is retained when a material change or a noteworthy update occurs, so counting rows is not the same as measuring how many hours each state lasted. Duration analysis would require rebuilding intervals between timestamps and defining the end of the observation period.

How to reproduce the results

The tables are generated directly from history.json. Each row supplies a state, operational category, confidence level and timestamp. The script groups values with counters, divides each group by the number of archived entries and rounds the share to one decimal place. The current condition and latest check are read from status.json.

The source files are publicly linked and can be downloaded without registration. The code does not insert hidden estimates or fill missing observations. If a row lacks a category, it is shown as unknown rather than assigned to the most likely group.

Limitations that must be respected

This report does not prove how many vessels crossed, how much oil moved or how long a physical interruption lasted. It also does not independently establish the accuracy of every news item. It is an editorial dataset describing decisions published by the monitor. Its value lies in traceability and in enabling outside review of the classification process.

Operational categories

Published confidence