Data rescue activities at Met Éireann (the Irish National Meteorological Service)

Met Eireann, Ireland's National Meteorological Service, maintains the National Climate Archive, a repository of paper and digital climate data dating from the 1800s to the present day. The archive contains an extensive collection of historical weather observations such as hourly and daily meteorological registers, daily and monthly rainfall registers and weather diaries.  In the last decade efforts have been made to image and transcribe the paper records to make it available for climate analyses.  The data rescue projects to date have involved collaboration with another government organisation, Irish universities and a citizen science project. These projects have involved manual data entry, which is a very time-consuming process, for example it takes approximately six hours to transcribe one month of daily climate data from one station.  As part of our efforts to enhance efficiency, we are currently evaluating the use of machine learning optical character recognition (ML-OCR).

The archives also hold paper charts such as barograms, anemograms, thermograms, pluviograms and hygrograms. The pluviograms are very important as they provide us with sub-daily rainfall measurements and vital information on historical high intensity short duration events. Although substantial manual effort has been undertaken to extract the information from the pluviograms, a considerable volume remains unrecovered. To address this, we have investigated a novel computer vision–based approach for converting the analogue data into digital records. This method is also applicable to other chart types.

This presentation will outline past data rescue projects that have been undertaken and the approaches taken to rescue the data. It will discuss current data rescue projects including Met Eireann's Irish Weather Rescue project-a citizen science project to rescue historical daily rainfall observations from stations across Ireland as well as research on the use of machine learning for transcribing climate observations and digitising charts.