First results from ROPEWALK (Rescuing Old data with People's Efforts: Weather and climate Archives from LogbooK records) - the digitization project for three centuries of weather observations on board of Danish ships

ROPEWALK, funded by the AP Møller Mærsk Fund, is a joint initiative of the Danish National Archive and the Danish Meteorological Institute, which aims at digitizing and transcribing all weather observations in Danish ship journals and logbooks stored in the Danish National Archive, consisting of more than 750 shelf metres beginning as early as the 1680s. With the exception of the Napoleonic wars and Danish state bankruptcy in 1814, the data is complete.

Ship journals over large parts of the Northern Hemisphere are kept in the archive, with two regions of particular interest, Greenland and the Øresund:

The Greenlandic Trade Company had a monopoly for commerce with the colony of Greenland for nearly 200 years, and foreign ships would not be allowed to call a port. The company conducted these "Greenland Voyages" to western Greenland several times per year, starting as early as 1721 and through the 1930s. Weather observations from these voyages often include detailed sea ice observations.

Every ship passing the sound or belts in Denmark had to pay for passage between 1426 and 1857. To ensure payment, Danish war ships were placed at strategic locations near Copenhagen, Helsingør and Nyborg. Weather observations on board of these ships go back to the end of the Little Ice Age. In several cases, observations were conducted every time the ship bell was struck, resulting in as many as 48 observations in the course of a single day. 

The scanning of the original logbooks and ship journals by the National Archive in highest possible resolution took 13 person-years, resulting in 2.1 million images covering more than 2.5 TB of data.  Up to roughly 1750, the data consists of diary-like daily note in free text. Starting in the 1710s, observations are recorded as numbers in preprinted tables. We have transcribed this latter dataset, constructed a data model, trained a machine learning algorithm and conducted quality control. Free text data will be considered later.

First results will be presented. All transcribed data will be made publicly available and can be used for future research or as input for reanalysis projects.