Ireland’s pre‐1940 daily rainfall records

This article presents daily rainfall data and metadata for Ireland transcribed from historical manuscript and printed copies of rainfall registers located in Met Éireann's archives. To facilitate the transcription of rainfall observations from paper records, the historical manuscripts were scanned and integrated into Met Éireann's digital archives. The transcription from digital image to data format was undertaken in collaboration with students at Maynooth University as part of a novel crowdsourcing initiative to integrate data rescue activities into the classroom. In total, 3,616 station years of rainfall data (~1.32 million daily values) were transcribed. The data, which was double keyed, have undergone basic quality assurance to check for transcription errors and the resultant raw data and associated metadata are presented here. Ongoing work involves the application of further quality assurance and homogenization techniques to develop a long‐term, quality assured daily rainfall network for Ireland.


| INTRODUCTION
There is an increasing demand for climatological information to support scientific understanding of the impacts of climate change. Accurate and reliable long-term meteorological records are essential for understanding climate trends and variability, evaluating reanalysis products and climate models and climate risk management (e.g. Matthews et al., 2014;Wilby et al., 2016;Noone et al., 2017;Murphy et al., 2019). Consequently, much work is being carried out internationally to rescue historical climate data and develop long-term climate series (e.g. Cornes et al., 2012;Allan et al., 2016;Ashcroft et al., 2018;Hawkins et al., 2019;Coll et al., 2019).
There is a long history of meteorological observation in Ireland. The earliest known weather observations were recorded at Kilkenny for the Duke of Ormond by John Kevan in 1682. From 1684 to 1686, William Molyneux and George Ashe, under the auspices of the Dublin Philosophical Society, documented weather at Trinity College, Dublin. Records resumed here in 1708 to 1709 when Samuel Molyneux documented observations of temperature, rainfall and pressure. In 1787, Richard Kirwan established the first Irish meteorological station at Cavendish Row, Dublin. These records are the earliest surviving observations compiled with the aid of accurate instruments. Kirwan maintained the observations until 1808, by which time regular meteorological observations had commenced at the National Botanic Gardens (Dixon, 1987). Observations from Armagh Observatory date from 1796 to present, providing the longest continuous series from any single site on the island (Butler et al., 2007). In 1829, the Ordnance Survey Office located in Phoenix Park, Dublin began recording systematic meteorological measurements providing continuous readings up to the present day.
In 1859, George James Symons, working with a network of voluntary observers, set up a system for gathering and publishing rainfall records from across Britain and Ireland. The data from Symons' observational network were published in British Rainfall (formerly English Rainfall), which for the year 1860 reports that there were then 168 stations in the network. No observations were included from sites in Scotland or Ireland, and only three from Wales (Glasspoole, 1952). The earliest available observations taken at Irish sites are from 1864. In 1900, the network founded by Symons became known as the British Rainfall Organisation and was later (in 1919) formally transferred to the aegis of the British Meteorological Office. At this time, there were approximately 5,000 observers contributing to the rainfall network (Glasspoole, 1952).
At the end of the 19th century, under the direction of the British Meteorological Department, led by Admiral Robert Fitzroy of the Royal Navy, an operational network of observing stations was established. In 1868, the Met Office equipped seven observatories, including two in Ireland, with autographic instruments, providing the first continuous record of a variety of meteorological parameters at a selection of locations (Eden, 2009). The records, along with the Phoenix Park records, which were maintained by the Royal Engineers, represent the early efforts of the British Meteorological Office to establish a network of meteorological stations across Britain and Ireland.
The Irish Meteorological Service, later Met Éireann, was established in December 1936 and subsequently took over responsibility for the network of Irish stations from the British Meteorological Office. The registers that pre-date the establishment of the Irish Meteorological Service were later transferred from the British Meteorological Office to the current service in Ireland. The substantial paper records are carefully preserved in Met Éireann's archives, but until now have largely remained in paper format for years prior to 1941.
Met Éireann maintains the National Climate Database. The database comprises observations received by Met Éireann from the current network of staffed, synoptic, climatological and rainfall stations. Observations are quality controlled and archived in Met Éireann's database. Daily observations of rainfall have been digitized back to 1941 and temperature to 1961 (Walsh, 2013). Work to digitize and transcribe long-term daily minimum and maximum air temperatures is currently underway at the National University of Ireland (NUI), Galway (Mateus et al., 2020). Other long-term series include work by Murphy et al. (2018) to compile a 305-year continuous monthly rainfall series for the island of Ireland (1711-2016) utilizing data from the UK and Ireland, including previously unpublished work from the 1970s by the British Meteorological Office. The comprehensive monthly series builds on the work of Noone et al. (2015) which developed a monthly rainfall series for 25 stations throughout Ireland for the period 1850-2015. However, until now there has not been a concerted effort to construct a long-term daily rainfall series for Ireland using historical records. This paper describes recent work, undertaken as part of the PhD research of the lead author, to digitize and transcribe historical daily rainfall records from Met Éireann's extensive archive collection. The main objective of this work was to create a digital archive of the paper records of Ireland's longest meteorological stations, and from these imaged records and additional rainfall registers to extend the availability of long-term daily rainfall data prior to 1941. The data presented here are the raw data and associated metadata. It is envisaged that by presenting the data in its original state it can be easily integrated into current international data rescue initiatives, for example Copernicus Climate Change Service Global Land and Marine Observations Database, and that future research will have recourse to the raw data.
The remainder of the paper is organized as follows: Section 2 provides a description of the data. First, we discuss the digitization of historical records from Met Éireann's archives; then, we outline the data transcription process and provide details of the stations transcribed, followed by a description of the metadata that was collected during the transcription process. In Section 3, we describe the process of checking for errors in the transcribed data. Section 4, provides details of dataset use, while Section 5 concludes with a proposed pathway to complete the imaging and transcription of stations and meteorological data that remain on paper.

| Data sources
The daily rainfall data were largely taken from annual Rainfall Registers and Meteorological Registers held in Met Éireann's archives. These registers are described in detail below. In addition, digital images of station records for Kells F I G U R E 1 Sample rainfall register. The annual forms report daily rainfall observations and station metadata for various locations throughout Ireland (Headfort) and the National University of Ireland (NUI), Galway were obtained from the National Library of Ireland and NUIG, respectively.

| Rainfall registers
Previous work by Met Éireann focused on imaging the single-sheet annual rainfall registers collected from the early rainfall network, specifically the original handwritten rainfall observations that were taken by the volunteers who worked as part of Symons' rainfall network in Ireland (Treanor et al., 2011). In 2014, the Met Éireann Library received funding to collate, catalogue and preserve the registers in this collection. The records are known as the 'Rainfall Registers' and contain daily rainfall data and station metadata from various locations throughout Ireland. Readings were taken once a day and recorded on a standard form issued by Symons and later the British Rainfall Organisation (Figure 1). The collection includes readings from every county in the Republic of Ireland and intermittent observations taken in Antrim, Armagh, Derry, Down, Fermanagh and Tyrone in Northern Ireland. There are over 700 stations in total, which return data for varying time periods between 1864 and 1940.

| Meteorological registers
Non-digitized records include a collection of meteorological registers consisting of 224 bound volumes, 128 folders and 353 folios of station records covering the period 1855-1976. The collection includes observations from Ireland's longest operating stations, many of which are still operational. The registers have been arranged by date and region and catalogued by their location in the archives at Met Éireann (Treanor et al., 2011). They provide a daily record of weather parameters at various locations throughout Ireland covering the mid-19th century until the latter part of the 20th century. F I G U R E 2 Sample meteorological register. The monthly forms report daily observations for multiple parameters including temperature, pressure, precipitation, sunshine, wind and other parameters Observers recorded data from a selection of instruments that measured temperature, pressure, precipitation, sunshine and other fields. Readings were taken at least once a day and recorded on a British Meteorological Office issued broadsheet form that was designed to contain a month's data ( Figure 2). At the end of the month, the forms were posted back to the British Meteorological Office for centralized calculations and corrections.
A Metis EDS professional digital scanner ( Figure 3) was purchased to progress the digitization of the historical meteorological registers. To date, meteorological observations from eight long-term stations have been scanned and integrated into the digital database (Table 1). Work is ongoing to digitize the remaining manuscripts. Individual pages were scanned as high quality tif files and categorized using the standard naming convention adopted by Met Éireann. The scanned images were then uploaded to Met Éireann's database where they will be made available for research purposes.

| Data transcription
Daily rainfall observed at 114 sites throughout Ireland was transcribed as part of this work ( Figure 4). Stations were selected based on record length, continuity and spatial distribution. Additional stations that could potentially be used to infill gaps in long-term data series were also included. The transcription from paper and digital image format to digital numerical format was largely undertaken by final year Geography students at Maynooth University as part of a novel crowdsourcing initiative to integrate data rescue activities into the classroom. Ryan et al. (2018) presented an innovative approach to data rescue by developing a research-led project to engage students in data rescue tasks for credit. The study explored (a) the potential for integrating data rescue activities into the classroom, (b) the ability of students to produce reliable transcriptions and (c) the achieved learning outcomes for students. The work was facilitated by the provision of student aids including written guidelines, transcription templates with an automated quality-assurance check, a video tutorial, in-class workshops and an online discussion forum. An evaluation of learning outcomes and student's perceptions of the project demonstrated a positive educational experience. Following the success of the initial F I G U R E 3 Metis EDS Gamma professional digital scanner used to image historical meteorological registers held in Met Éireann  T A B L E 2 (Continued) project, a further two iterations have been executed across three cohorts of final year Geography students at Maynooth University, producing in excess of 3,500 station years of historical daily rainfall data. A detailed description of the methodology and access to resources is provided by Ryan et al. (2018) (https://doi.org/10.1175/BAMS-D-17-0147.1). The majority of the data was transcribed from the annual rainfall registers. Additional rainfall data were extracted from the long-term meteorological registers listed in Table 1. Rainfall register observations were, for the most part, recorded in inches of rainfall. Data extracted from the meteorological registers were recorded in millimetres from the start of record up to 1914/15 when the unit of measurement changed to inches. Table 2 Table 2.

| Metadata
In addition to the rainfall observations, metadata for each station were extracted from the individual station records and recorded in separate text files ( Figure 6). These files provide information on station name and location, observer, record length, missing data, diameter of gauge, changes in gauge height, gauge pattern, time of observation, unit of measurement and distance to nearest railway station. A section for 'Additional Information' provides details of observations made while transcribing the data, for instance, the presence of a multiday accumulation. A separate section for 'Notes' provides a transcription of any handwritten notes recorded on the original record by the observer. For example, an observer may note an exceptional rainfall event or a leak in the gauge. These notes are transcribed verbatim.
The British Meteorological Office coordinated the meteorological network by supplying instruments and instructions for observations. Nevertheless, before standard equipment and procedures for meteorological observations were introduced from the late-nineteenth century to mid-twentieth century, the design and placement of rain-gauges varied considerably. Observer bias, instrument changes, sampling periods, as well as external factors relating to site exposure (e.g. proximity to buildings) effect the accuracy and consistency of observations (Kunkel et al., 2005;Daly et al., 2007;Green et al., 2008). In general, such changes were recorded on the station record and subsequently transcribed to the metadata files. Detailed metadata presented here provide the comprehensive account of station changes and can be used as an aid in determining the reliability of station records. This information will be particularly valuable in helping to explain the presence of any abrupt shifts identified when further quality assurance and homogenization techniques are applied (Aguilar et al., 2003).

| ERROR CHECKING
At each stage of the transcription process, quality assurance measures were employed to preserve the integrity of the data being rescued. Keying guidelines were developed ensuring conformity to World Meteorological Organisation (WMO) standards (WMO, 2016). Monthly totals were examined against the derived sum of the daily entries to identify potentially incorrect data entries. The data were double keyed and the entries from different transcribers compared.

F I G U R E 5 Number of stations each
year for which daily rainfall data was transcribed. Post-1940 data are available through Met Éireann's website Where the entries agreed, the value was provisionally accepted as the raw data value. If the values disagreed, the original record was manually examined to ascertain the true observed value. An examination of errors across all transcriptions revealed a percentage error of <1%. Multiday accumulations were identified and flagged using the original records as a reference. A description of numerical flag values is included in the metadata files. These indicator flags will facilitate the re-distribution of multiday accumulations to the respective days on which no observation was recorded. This will be undertaken prior to the application of further, more sophisticated, quality-assurance techniques. As a final check for transcription errors, the upper and lower 1% of observations (non-zero precipitation) were examined for each individual station record. Values identified as outliers were cross-checked against the original record.

| DATASET USE
The data are freely available from the edepositIreland data centre (http://hdl.handle.net/2262/91347). The dataset comprises daily rainfall data for 114 stations at various locations throughout Ireland for varying time periods. Individual station folders contain two files: a data file in ASCII format and a corresponding metadata text file as described in section 2.4. Data files consist of five columns providing the observation date (year, month and day), followed by rain value and indicator value. The indicator value provides information about the nature of the observation and identifies multiday accumulations.
A key to the respective indicator values is provided in the metadata files. Rainfall values run continuously from start date to end date of the data recovery period, with missing values denoted using a −999 indicator. Work is currently underway to F I G U R E 6 Sample metadata file providing information regarding station name and location, observer, record length, missing data, diameter of gauge, changes in gauge height, gauge pattern, time of observation, unit of measurement and distance to nearest railway station. A section for 'Additional Information' provides details of observations made while transcribing the data, for instance, the presence of a multiday accumulation. A separate section for 'Notes' provides a transcription of any handwritten notes recorded on the original record by the observer produce a network of long term, quality assured daily rainfall stations using the datasets whose generation is described in this paper. Preliminary quality assurance checks have been applied to assess the accuracy of the transcription process. A second, comprehensive set of quality assurance techniques will be applied to detect both systematic and non-systematic errors, this will be described in a subsequent paper. Post-1940 daily rainfall records are readily available from Met Éireann's climate database. The newly transcribed data will be added to the database and used to extend these station records back to the late 19th Century. Once joined, the full series will be homogenized and analysed to assess variability and changes in the characteristics of rainfall events over the long-term record.

RESCUE
The importance of historical climate data is being increasingly acknowledged for its role in supporting effective climate risk management through reanalysis and validation of climate models. As a consequence, climate data rescue has experienced a substantial rejuvenation in recent years, with a number of national and international projects underway, for example Copernicus Climate Change Data Rescue Service (C3S) and the International Data Rescue (I-DARE). The data presented here mark a significant and innovative effort to progress data rescue efforts for Ireland. Nevertheless, a considerable amount of data has not yet been digitized and exist only in hard copy format. Met Éireann, as part of their current operational business plans, has considered opportunities to advance data rescue initiatives within the organization and through collaboration with other agencies to create a comprehensive climate data bank and to facilitate the creation of high-quality data products.
Met Éireann are involved in a number of ongoing projects and initiatives aimed at enhancing climate data availability and accessibility. These include a collaboration with The Central Statistics Office (CSO) to transcribe all parameters and metadata from the eight meteorological registers imaged during the course of this work (see Table 2 for station details). As part of this, transcription of the entire Phoenix Park series has recently been completed. The data series which spans the period 1829 to present is the longest continuous series for the Republic of Ireland and the second longest in Ireland after Armagh Observatory. The extent of the data available for the Phoenix Park makes it one of the most comprehensive series available worldwide.
Engagement of non-experts or 'citizen scientists' on a voluntary basis has become increasingly significant to the rescue and refinement of observational data across multiple scientific disciplines (Bonney et al., 2014). The success of ongoing citizen science applications -for example, OldWeather. org (www.oldwe ather.org/), the Weather Rescue Project (https://weath erres cue.wordp ress.com/) and Data.Rescue@ Home (http://www.data-rescu e-at-home.org) underscores the potential of crowdsourcing as a data rescue strategy. Further, the data rescue project developed by Ryan et al., (2018) as a collaboration between Met Éireann and Maynooth University will continue as an integral part of the Climate Change module delivered by Dr. Murphy. It is hoped that the project will be developed further through the use of an online platform designed to host Ireland's historical meteorological records. Such an application would facilitate the extension of the data rescue project to other teaching programmes and significantly advance contributions from citizen scientists.
Climate data rescue must be viewed as a continuous, long-term activity (Brönnimann et al., 2018). The priority for Met Éireann over the coming years is to catalogue and image all historical records currently held in the archives for integration into the climate repository, an effort enhanced by the development of resources (including a digital scanner) as part of this work. Continued effort is also being made to recover records held in other libraries and institutes in Ireland and abroad. Met Éireann have recently recruited an archivist to document the vast historical record holdings including meteorological and climatological collections, weather diaries, monthly bulletins, annual reports, weather maps and site inspection reports. Providing access to these records, along with an inventory of what is available is important to promote collaborations and knowledge sharing between interested stakeholders.