(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Digitizing historical daily weather bulletins through citizen scientists: The ReData project [1]

['Alessandro Ceppi', 'Department Of Civil', 'Environmental Engineering', 'Dica', 'Politecnico Di Milano', 'Milano', 'Associazione Meteonetwork Odv', 'Veronica Manara', 'Department Of Environmental Science', 'Policy']

Date: 2026-04 

In recent decades, numerous climate data rescue programs have begun in many countries worldwide. These projects aim to preserve data recorded on paper sheets, which are vulnerable to deterioration, and to make them accessible to the scientific community. This enhances the accuracy of climatological studies and historical reconstructions, including those focused on specific events. This study presents the designed framework developed in the ReData (Recovery of Data) project, launched by the Meteonetwork association in collaboration with the University of Milan in 2017, and upgraded in 2024 on the Zooniverse platform, which engages volunteers in scientific research activities; in particular, it showcases the methodology implemented for this climate data rescue initiative. The job leverages the potential of citizen science to digitize meteorological data, collected by the Italian Royal Central Meteorological Office (RCMO) from 1879 to 1940 and published in daily meteorological bulletins, on a platform specifically designed to facilitate large-scale digitization and ensure user accessibility. In addition, as a practical application of digitized data, a case study is presented involving a synoptic reconstruction of the flood event that affected the River Adige in northeastern Italy in September 1882. The overall project provides critical data for reanalysis models, and it enhances the understanding of historical climate trends over the Italian peninsula, offering significant cultural and scientific value.

Funding: This work has been financed by Research Funds from the Italian Ministry for University and Research (PRIN 2022 – CN4RWK–CCHP-ALPS Climate Change and Hydro-Power in the Alps, funded by the European Union (Programme Next Generation EU)).

Copyright: © 2026 Ceppi et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

The main goal of this paper is to highlight the methodology we adopted to engage citizen scientists for the data digitization activities carried on in the ReData project. The text is structured as follows: Section 2 describes the database, i.e., the daily weather bulletins edited by the RCMO. The core of the paper is presented in Section 3, which details how we organized the ReData digitization on the Zooniverse platform and outlines key challenges encountered in this activity. Lastly, a reconstruction based on sea level pressure data related to a severe flood that occurred in the Adige River basin is illustrated in Section 4, while discussion and conclusions are reported in Section 5.

Based on a detailed analysis of the evolution of the RCMO weather bulletins, regarding both monitoring stations and recorded meteorological variables, the aim of this study is to introduce the implemented design and describe the methods for this activity, which is structured to streamline the digitization process for citizen scientists, even replicable by other researchers. The final goal is making freely accessible all the information contained in the RCMO weather bulletins to support and contribute to scientific research. In order to better highlight the potential of rescued data, a case study is presented, regarding an analysis of a severe flood event occurring in September 1882 in northeastern Italy, which, due to persistent heavy rainfall, caused huge damages, including the collapse of Verona’s Ponte Nuovo bridge. The weather data from RCMO bulletins showed detailed information at synoptic scale over the Italian peninsula, demonstrating their added value in the meteorological event reconstruction.

The bulletins, issued by the Meteorological Office and scanned within the ReData project, report the weather analysis daily performed. They include everyday data for many Italian and some European stations as well as weather charts displaying the spatial distribution of the most important atmospheric variables. Therefore, one of the main tasks of the ReData project, which is currently ongoing, is the digitization of the huge amount of data contained in these bulletins.

In the ReData project, we scanned all the data contained in the daily weather bulletins issued by the Italian Royal Central Meteorological Office (RCMO) between December 1879 and April 1940. This Meteorological Office was established on the 26 th of November 1876 (Royal Decree n. 3534), some years after the unification of the Kingdom of Italy (1861), and it played a pivotal role in coordinating meteorological observations across the country for a long time. It evolved from earlier initiatives at the Collegio Romano in Rome, where meteorological data were originally collected to support astronomical research, and set the basis to unify Italy’s fragmented meteorological networks by standardizing measurements and enabling consistent data collection nationwide. Over time, the office’s responsibilities expanded to include geodynamics, later evolving into the Central Office of Meteorology and Geophysics. Despite challenges, such as limited resources and shifting institutional priorities, the RCMO laid the groundwork for Italy’s modern meteorological and seismic services. In 1925 the Italian government decided to move the forecasting activities to the Italian Air Force. The RCMO, therefore, shifted its focus toward climatology and agricultural meteorology, and the archives of this institution are currently part of the Italian Agriculture and Environment Research Centre (CREA).

Citizen science involves the active participation of the public in scientific research, allowing individuals without formal scientific training to contribute to data collection, analysis, and problem-solving. This approach enables researchers to gather large volumes of data while fostering greater public engagement with scientific issues [ 23 – 25 ]. Organizations such as the Association for Advancing Participatory Sciences (AAPS, https://participatorysciences.org/ , accessed on 1 st July, 2025) and the European Citizen Science Association (ECSA, https://www.ecsa.ngo/ , accessed on 1 st July, 2025) have been established to support and promote these initiatives, with this latter providing a framework of key principles for best practices. Indeed, the digital age has significantly advanced citizen science, with dashboards like Zooniverse ( https://www.zooniverse.org/ , accessed on 1 st July, 2025) enabling global collaboration.

Under this framework, the proposed study focuses on a recent project regarding the recovery of historical climate data in Italy: ReData ( https://datarescue.climate.copernicus.eu/redata accessed on 1 st July, 2025). The project is led by the Meteonetwork (MNW) association [ 22 ] in collaboration with the University of Milan (UniMi). Initially launched in 2017 on MNW webpage ( https://www.meteonetwork.it/ , accessed on 1 st July, 2025), it was later upgraded in 2024 and moved to the Zooniverse platform ( www.zooniverse.org/projects/meteonetwork/redata , accessed on 1 st July, 2025) to benefit from citizen scientists’ contribution.

In Italy, a country with a rich heritage of historical meteorological data, home to the invention of famous instruments such as Galileo’s thermoscope and Torricelli’s barometer, and creator of the first international observational network by the Accademia del Cimento in the 17 th century [ 18 ], the Italian Association of Atmospheric Sciences and Meteorology (AISAM) plays a central role in national data rescue efforts. In 2022, AISAM launched the Cli-DaRe (Citizen Science for Italian Climate Data Rescue) project, aiming to initiate a long-term program that leverages the vast potential of citizen science to recover the extensive meteorological data still preserved in Italian paper archives. The first initiative under this project, Cli-Dare@School, involved high school students in the recovery of historical rainfall series from monographs published by the Italian Hydrographic Service between 1918 and 1966 [ 19 – 21 ]. Other projects are ongoing, and further information can be found at the following website: https://aisam.eu/progetti-aisam/ (accessed on 1 st July, 2025).

Another indicator of the growing importance of climate data preservation within today’s scientific community is the emergence of numerous non-profit organizations dedicated to promoting and raising awareness about this issue. A remarkable example is the International Environmental Data Rescue Organization (IEDRO), a U.S.-based non-profit society that collaborates with the WMO, the National Oceanic and Atmospheric Administration (NOAA), and the National Meteorological and Hydrological Services (NMHS) of developing countries to support data recovery efforts. Lastly, other worthy examples come from the potential of enrolling undergraduate students in climate data rescue projects [ 14 – 16 ], even for solving metadata matters [ 17 ].

One of the largest climate-related citizen science data rescue projects ever completed comes from the Rainfall Rescue project [ 12 ] where more than 66,000 paper sheets containing 5.28 million handwritten monthly rainfall observations recorded across the UK and Ireland between 1677 and 1960 were digitized. This achievement was made possible by the contributions of over 16,000 volunteers who participated during the early months of the COVID-19 lockdown in 2020 through the Zooniverse platform. In New Zealand, the Southern Weather Discovery is another remarkable initiative focused on transcribing meteorological records from thousands of ships that crossed the Southern Ocean during the 19 th and 20 th centuries to improve the understanding of historical weather patterns and assess long-term climate trends [ 13 ]. In France, the national meteorological service Météo-France, in collaboration with the National Archives and supported by the BNP Paribas Foundation, has been engaged since 2012 in a program to preserve climate archives from both France and its former colonies; covering the period from 1850 to 1960 ( http://archivesduclimat.meteofrance.fr/ , accessed on 1 st July, 2025).

The preservation of historical meteorological data has been ongoing for several decades, such as the International Surface Temperature Initiative [ 11 ], but its progress varies depending on the region, accessibility of sources, and availability of resources. Several projects have already been completed, and others still need to be initiated to safeguard data that have not been digitized yet. At international level, the World Meteorological Organization (WMO) coordinates global efforts to standardize and digitize climate data, creating databases that are accessible to scientists worldwide. For instance, the Data Rescue Portal, by the Copernicus Climate Change Service (C3S), collects and disseminates information on past, current, and planned data rescue projects ( https://datarescue.climate.copernicus.eu/ , accessed on 1 st July, 2025).

Unfortunately, many old meteorological data, originally recorded on paper documents, have yet to be recovered, and their preservation is essential for several reasons [ 9 ]. First, these paper records are often significantly deteriorated, posing a serious risk of permanently losing valuable information; moreover, making these data accessible for climatological research would enhance the accuracy as well as the spatial and temporal coverage of datasets used to investigate the climate of the past. The recovery of new data can therefore allow researchers, on one hand, to improve the climate reconstruction at local scale and, on the other hand, it can contribute to increasing the accuracy of global reanalysis datasets [ 10 ].

Climate change is one of the most urgent and complex challenges in the last decades [ 1 ]. Historical data allow scientists to detect long-term trends, identify anomalies, improve reanalysis and validation of climate models, which are fundamental for developing effective mitigation and adaptation strategies. Trends in meteorological variables can be studied by investigating records of observational data at selected station sites (see, e.g., [ 2 ] for Italy or [ 3 ] for the Greater Alpine Region) as well as by considering gridded datasets of observational records (see, e.g., [ 4 ] for Europe, [ 5 ] worldwide). In addition, global reanalysis datasets, such as the ECMWF (European Centre for Medium-Range Weather Forecasts) Reanalysis v5 (ERA5, [ 6 ]) and the NOAA Twentieth Century Reanalysis (20CR) [ 7 ], play a key role in reconstructing climate of the past by assimilating old observations into state-of-the-art dynamic models [ 8 ]. However, both for gridded datasets of observational data and for reanalyses, the reliability of reconstructed fields directly depends on availability, homogeneity, and spatial distribution of the underlying measurements.

In 1914, evening observations from the day prior to the bulletin’s issue date were added too. In the following years, new observations were gradually introduced, primarily to provide a more comprehensive set of variables to meet the needs of aviation, which became the bulletin’s main scope in 1927.

The most relevant section for the current objectives of the ReData project is the table reporting measurements from Italian stations. In some periods, this table is reported onto a single page; in others, it spans over two or three pages. While this table underwent changes over the years, it kept a stable structure during the period from 1882 to 1913 where the following meteorological variables were recorded every morning ( Fig 3 ): barometric pressure at sea level, 24-hour barometric tendency, air temperature, 24-hour air temperature tendency, wind speed and direction, sky condition, sea state, cloud coverage, 24-hour rainfall, daily minimum and maximum temperatures, miscellaneous observations (typically, a brief comment on weather events over the previous 24 hours), and notes.

The daily weather bulletins of the Italian RCMO are preserved in various archives, either as separated files or compiled into annual volumes. Although relevant changes occurred over time, the general structure remained largely consistent, and it can be divided into four main sections: (i) a table of meteorological observations from Italian stations, (ii) a table of meteorological observations from stations in other European countries, (iii) one or more weather charts, (iv) other information, including a short-term weather forecast.

In addition to the number of reported Italian stations, many other changes occurred in the bulletin over the 60-year period during which it was issued. These changes concern the format and organization of the bulletin, the number of pages, the reported variables and weather charts, the information (station data and synthetic reports) from other European countries (not reported in Fig 1 ), and the format of short-term weather forecasts, referred to as presagi, which was reported on the last page. More detailed information on the temporal evolution of the daily weather bulletin issued by the Italian RCMO is provided in Fig 2 .

In 1879, the RCMO established a national network for the collection of near real-time observations. Using telegraph technology, the station observations were transmitted to the Central Office in Rome, enabling the publication of a daily weather bulletin. At the beginning, the network included 10 Italian stations, which quickly increased, reaching about 70 within a few decades. Starting from 1914, it further expanded to also include, beside the traditional network of meteorological observatories mainly located in Italian large cities, stations of harbor authorities and cities in newly acquired territories, such as Trento and Trieste, as well as colonial locations like Tripoli and Benghazi in Libya, amounting to more than 100 stations ( Fig 1 ). Over the years, the number of stations showed significant variability, with a decrease after 1922; nevertheless, the station density remained high enough to meet the requirements for synoptic meteorology.

3 ReData activities

3.1 Scanning the RCMO weather bulletins The first achievement of the ReData project was the scanning of the daily RCMO weather bulletins which are now freely available as JPG files (https://doi.org/10.13130/RD_UNIMI/R1GVKF - [26]) from December 1879 to April 1940, totaling over 60 years of data and comprising more than 22,000 bulletins. As each bulletin was scanned page by page, the collection amounts to 99,518 individual scans, with formats ranging from 23 × 31 cm for the oldest bulletins to 35 × 50 cm for the most recent ones. Most of the scans (about 95%) were made from bulletins stored at the Department of Environmental Science and Policy of UniMi, while the remaining scans were obtained from bulletins preserved at the Central Historical Library of the Collegio Romano in Rome. The scans of the bulletins from UniMi were carried out between 2016 and 2025 by the Garon private company; the remaining scans were completed in February 2025 as part of the “Dieci e Lode” project [27], funded by the Italian Ministry of Culture (https://aisam.eu/progetti/10-e-lode-en/, accessed on 1st July, 2025). To scan all sheet documents with dimensions equal to, or smaller than, A3 format, we used a Xerox flatbed scanner at a resolution of 300 dpi, calibrated to the original document size, while for documents larger than A3, we employed a portable L-shaped scanner, the Czur Shine Ultra Pro, equipped with a 24-megapixel sensor. Images digitized using the flatbed scanner did not exhibit significant quality issues, except in cases where the original documents were already damaged; however, a nominal resolution of 300 dpi was consistently maintained for all scans performed with this tool. On the contrary, scans performed using the portable L-shaped scanner, employed for documents from 1927 onward, exhibited some variability in nominal resolution due to their larger format. This variability is attributable to differences in operational setup, including original document size, scanning distance, and cropping procedures. Nonetheless, the effective resolution remained close to 300 dpi, with a slight decrease observed for larger format documents. The originals from the UniMi archive, preserved as loose sheets, were digitized without significant geometric distortions. In contrast, the volumes from the Collegio Romano archive which include previously missing years and recovered individual days, are bound in particularly compact folders. This binding made it difficult to scan the documents on a flat surface, resulting in noticeable distortions, especially near the spine. Such distortions may hinder subsequent optical character recognition (OCR) processes, even when using artificial intelligence-based tools.

3.2 First attempt to digitize data A first attempt to digitize the data in the RCMO weather bulletins was carried out using a web application developed by MNW, where users were entirely free to settle what to digitize, following a vertical approach, that is, once a city was selected, they could choose a time period spanning multiple available years, with each daily digitization corresponding to the completion of several data rows. The idea behind this method was to encourage engaged citizen scientists to adopt a specific monitoring station, enabling them to consistently work on the same location, and thereby contribute to building long-term records of various meteorological variables reported in the bulletins. The results obtained using the MNW application were initially very encouraging: in fact, from the platform’s launch in 2017, a significant interest was observed, with the number of rows entered per day peaking at nearly 300. Data entry remained relatively stable until July 2019, with spikes corresponding to the addition of new scans to the platform. Activity began to decline thereafter, with a temporary resurgence during the COVID-19 lockdown period. However, after September 2020, contributions significantly dropped. Over the entire lifespan of the platform (which remains available, but no longer updated or actively maintained), a total of 21,287 observation rows were digitized, covering 39 cities and involving 88 users. The distribution of entries by city shows that, in the cases of Verona and Milan, user participation enabled the digitization of a substantial volume of data for the period 1879–1909. In particular, Verona saw the near-complete digitization of all available information, with a total of 7,520 rows entered, while Milan reached 4,590 digitized daily records. From this experience, we derived three key take-home messages: (i) without the ability to engage a large number of potential collaborators, it is not feasible to digitize such a vast dataset; (ii) the vertical approach adopted was not the most effective way to exploit this type of dataset; and (iii) it is unrealistic to expect digitizing everything at once; therefore, it is essential to establish a prioritization strategy that focuses first on what is really possible to digitize. Based on these considerations, we opted for a new design of the digitization activity.

3.3 Setting up the data digitization on the Zooniverse platform Zooniverse is an international platform that hosts projects across a range of disciplines, including astronomy, biology, and climatology. It provides volunteers with simple explanations, tutorials, and forums to guide their participation. Since its launch in 2009, it has engaged over 2.8 million volunteers, completing more than 900 million classifications across 95 active projects (https://www.zooniverse.org/projects, accessed on 1st January 2026). Indeed, Zooniverse, which has already leveraged citizen science to successfully support historical meteorological and hydrological data recovery initiatives, such the monthly rainfall observations taken in UK [12] and SIREN (Saving Italian hydrological Measurements) [28] projects respectively, offers a valuable opportunity for ReData. The main feature of the projects on the Zooniverse platform is that they do not require specialized scientific knowledge, but involve processing large datasets and tasks that would be infeasible for a small group of researchers to handle individually. Building on what we learnt from the digitization of the RCMO weather bulletins on the MNW platform, we decided to move our activity to the Zooniverse platform in 2024 and to design the digitization in order to (i) focus on pages containing the data of Italian observatories only, (ii) consider, at least in a first moment, a subset of the observatories (about 40) that can provide a good description of the spatial coverage of the reported variables over the Italian territory, (iii) adopt an horizontal approach focusing on one year at time which gives priority to produce spatial fields rather than single station long-term records, (iv) limit the digitization to the period 1882–1939, as prior to 1882 the bulletin underwent too many changes, and in 1940 bulletins were available only for the first four months and were issued in a reduced form. Overall, the total number of considered years is 58 including 21,183 bulletins. Since bulletins are scanned page by page, the amount corresponds to over 96,228 scans. Considering the number of meteorological variables recorded, the project estimates over 25 million individual data points that could be digitized through ReData. Setting up the digitization project, the first task was to associate each scanned page with the corresponding observatories reported. Since the list of stations on each page remained consistent within the bulletins of a given year, these associations were established based on the first day of the year only. Accordingly, for each year, all the first pages of the daily bulletins were linked to a list of stations, all the second pages to another list, and so on. In addition, as Zooniverse does not accept files larger than 1 MB, it was necessary to slightly reduce the quality of any files exceeding this size, without affecting the document readability. An alternative option was resizing the images by cropping the edges, but this method was discarded, because the scans are often skewed, and it was not feasible to individually check each scan, with a real risk of cutting off useful content. The second task was defining the digitization workflows. For each year, a list of workflows was created, matching the number of municipalities cataloged in the bulletins. Each workflow is named after a specific city, and it groups all sheets containing information for that city. This structure allows for the digitization of all station-related data for a given year. The third task involved defining the set of stations to be included in the digitization activities. We decided to consider approximately 40 cities listed in the bulletin up to 1893. For the subsequent period, although data availability in the bulletin significantly increases, we are going to maintain the same number of stations, adding new ones only to continue the records of stations that ceased to be reported or to cover areas that were previously unrepresented (e.g., stations in former Italian colonies). The fourth task regards the variables to be included in the digitization activities. We adopted the same approach used for station selection: all variables reported up to 1913 were digitized, and, where possible, the same set of variables will be selected for the subsequent years. In setting up the digitization project, we chose to digitize all data from scratch, even though a small portion had already been digitized using the MNW application. This decision was made because the previously digitized portion was relatively small and, in that earlier effort, each data entry was digitized only once, whereas the new digitization protocol requires each entry to be digitized three times [9]. Nevertheless, the dataset produced by MNW will still be useful for cross-checking the newly digitized data.

3.4 Ethics Statement This publication uses data generated via the Zooniverse.org platform under the Zooniverse User Agreement and Privacy Policy (https://www.zooniverse.org/privacy). The authors do not have access to any personally identifiable information about the volunteers.

3.5 ReData on Zooniverse The current section outlines the structure of the ReData project on the Zooniverse platform that enables users contributing to data digitization in a simple and intuitive way, without requiring explicit technical or scientific skills. The homepage (Fig 4) includes a navigation menu, a blue banner for displaying announcements, a brief overview of the project, a list of active workflows, and user classification statistics. From the menu, users can access the following sections: (i) more information about the project, including details on the research team, achieved results, and frequently asked questions (FAQs) (via About tab); (ii) a randomly selected classification task from the active workflows (via Classify); (iii) a discussion forum (via Talk) where users can interact with other volunteers and the research team; (iv) the broader Zooniverse collections (via Collect); and (v) the user’s recent classification activity (via Recents). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. Overview of the Re-Data project homepage on Zooniverse ( Overview of the Re-Data project homepage on Zooniverse ( https://www.zooniverse.org/projects/meteonetwork/redata , accessed on 1st January 2026) with the main features of the platform. https://doi.org/10.1371/journal.pclm.0000865.g004 The Talk section is particularly important for the project’s success, as it allows volunteers to share questions or suggestions about bulletins and workflows, and to report potential errors or inconsistencies. This continuous interaction has enabled us to refine the data collection process and ensure higher data quality. In the central part of the homepage, users can select a specific workflow, identified by station name and year. However, the individual classification task (i.e., the specific day to be digitized) is randomly selected from the available entries within that workflow. Once a workflow is selected, either from the central page or the Classify menu, a scanned image from the bulletin appears to the volunteers. Then, they complete a series of tasks to digitize all variables for the specified station, year, and day, thereby completing the classification. To accommodate user preferences, all workflows for a given year are made available simultaneously, allowing volunteers to choose based on their interests, whether a coastal city, a major urban area, or a specific region. As previously mentioned, we upload one year at a time. This approach allows us to showcase project progress gradually, which we have found to be more motivating for volunteers than waiting until the entire project is completed. Additionally, it enables us to download and inspect data incrementally, helping to monitor and ensure the quality of the digitized information. The required sequence of tasks (Fig 5), which allows a classification to be completed, follows the layout of the observation table, proceeding from left to right across a row. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 5. List of the tasks that are necessary to fill in in order to complete the classification. https://doi.org/10.1371/journal.pclm.0000865.g005 Every task includes a description of variables to be digitized, a data entry field (either a text box or a dropdown menu), and a link to help documentation to assist users. The tasks are seven. The first allows us to check whether the selected station is actually present in the scanned image, the second to add a line in the image of the scanned page to avoid confusing rows in the tables. The remaining five allow to insert the following information: sea level pressure (SLP) and corresponding 24 h tendency;

temperature and corresponding 24 h tendency;

wind direction and speed;

sky condition, clouds direction and sea state;

maximum and minimum temperature and total precipitation in the last 24 hours. The organization of these tasks has slightly changed from when we launched the ReData project on Zooniverse thanks to the engaged users’ suggestions which helped us to optimize it. Moreover, moving on the right of the page, during the digitization it is possible to open a field guide (selecting Field Guide) where volunteers can find information about the history of the RCMO weather bulletins, an explanation on how to interpret some symbols during the digitization and the main FAQs. In particular, the Field Guide focuses on providing in-depth knowledge about the bulletins, including insights and curiosities related to the project. While optional, it enhances volunteer engagement by broadening their understanding of the project’s significance. Another essential section of the platform includes the tutorial, which is available during all steps of the digitization (selecting Tutorial). It provides a brief walkthrough of the project’s goals and tasks, guiding volunteers on how to perform classifications.

3.6 Beta review process Before the official launch, the project underwent a two-stage review process. The first stage, an internal review, involved a thorough evaluation by Zooniverse team members. During this phase, the project structure and page content were assessed to ensure they met specific standards, and detailed feedback was provided, including suggestions for potential improvements. The second stage was the beta testing phase, where the project was introduced to a limited group of Zooniverse volunteers, which participated by submitting beta classifications and sharing their comments via a standardized Google form. These phases were critical as they allowed for ultimate adjustments before the final launch on the platform, and they also tested data aggregation and analysis methods designed for the project. For beta testing, a reduced subject set was required to effectively assess the workflows during this period. Since these workflows planned for the ReData project were identical, only differing by station they referred to, a single workflow was proposed for the beta phase (Belluno 1882). A reduced subject set of 181 daily scans, created from January to June 1882 (six months in total), was associated with it. This testing phase provided an opportunity to gauge the responsiveness of the Zooniverse community: over 900 classifications were completed in less than two weeks. In this initial phase of the project, the discussion board proved to be highly valuable, as volunteers shared questions and observations that were instrumental in improving the site. The project was officially launched on 26th November, 2024.

3.7 Key challenges Although the primary activity of the project is straightforward and schematic, and volunteers are simply required to carefully copy data recorded in the bulletins, there are some intrinsic challenges presented by documents, paramount to this project. Firstly, handwriting poses significant encounters. Once users select a workflow, they must verify that the station (or city) name, written in an old italic style, appears in the first column of the bulletin. Then, users must transcribe a large number of values. At first glance, this may appear simple, but it is not always clear-cut. For instance, the same digit is sometimes written in different styles, which may introduce uncertainty. The primary challenges identified so far involve the numbers 5 and 7. Many of these ambiguities have been solved by cross-referencing the second and fourth columns of the bulletin. These columns contain information on the 24-hour differences in pressure and temperature, which link data of a specific day to that of the previous day; through such cross-checking, it has been possible to accurately decipher unclear handwriting. In certain tasks, such as identifying wind and cloud directions, users are asked to recognize an acronym in uppercase letters representing one of the 16 cardinal points. Since the response options are limited in these cases, a dropdown menu is added to facilitate the process. In other tasks, users must recognize words, often adjectives describing sky, sea, and wind. Many of these terms are frequently repetitive, thus, a dropdown menu is implemented as well. However, for non-Italian-speakers, decoding certain words can be difficult, particularly when abbreviations are used (e.g., ab. forte, deb.). Some examples are provided in Fig 6. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 6. Examples of how to decipher calligraphy for numbers, words and acronyms taken from the Field Guide, which volunteers can access at all times. https://doi.org/10.1371/journal.pclm.0000865.g006 This use of abbreviations progressively increased as bulletins became more detailed, first in 1923, and then in 1927. In some cases, reading these new entries can be quite tough (Fig 7). However, the community has adapted well, recognizing that the correct approach is simply to transcribe what they see. With fewer than ten distinct descriptions in regular use, this issue has become less problematic over time. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 7. Comparison between different yearly bulletins in terms of density and use of abbreviations. https://doi.org/10.1371/journal.pclm.0000865.g007 Lastly, one of the most persistent challenges is the quality of the original documents. Although the scans themselves are high-resolution, the source materials of more than a century ago often feature faded ink and occasional degradation due to poor storage. Combined with the ornate handwriting styles of the period, these factors pose relevant obstacles. Unfortunately, we cannot resolve these issues, but we have done our best to mitigate all other concerns.

3.8 The first five months on Zooniverse: project progress and community engagement Six months after its launch, the digitization of the first four years (1882–1885) was completed at a rate of approximately one year every five-six weeks (data updated to 15th May, 2025). Each owner of the project can directly download data from the Zooniverse page. These are included in one file saved in CSV format, where each row represents a classification (i.e., a day for one station with all available variables), and each column corresponds to a task saved as a string in JSON format. In addition to the digitized data, a substantial number of metadata is available for all classifications, allowing to monitor the progress of the project while preserving the anonymity of the volunteers. Specifically, we digitized 34 stations for the year 1882 and 40 stations for the subsequent years, resulting in a total of 171,437 classifications. At first, we checked the data to ascertain whether all the expected days for each station and year were included in the dataset, verifying the correct working of the platform. For our project, we required three classifications per day; however, we observed that if a workflow is not retired immediately upon completion, it remains available and may be classified again. Therefore, careful monitoring of classification progress is necessary to ensure workflows are promptly retired; otherwise, it sometimes occurs that volunteers classify the same day multiple times. By the end of this analyzed period, thanks to the internationality of the Zooniverse platform, 1,153 volunteers completed at least one classification, with participation increasing almost linearly. After an initial peak of 303 volunteers, the number of active participants per week stabilized between 50 and 100 (Fig 8A), ensuring a weekly mean of 4,762 classifications. The number of daily classifications (Fig 8B) was highly variable and, as expected, followed a weekly cycle, with higher values during weekends and holidays (e.g., Easter). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 8. A) Weekly number of active volunteers; B) Daily number of completed classifications; C) Weekly median time required to complete a classification; D) Percentage of classifications for each time interval. The information refers to the period between 26th November, 2024 and 15th May, 2025. https://doi.org/10.1371/journal.pclm.0000865.g008 An analysis of the classifications grouped by username shows that 47.8% of all classifications were completed by three volunteers, and 29.4% were performed by a single “Top Classifier” who systematically catalogued all available data. When new workflows are activated, the Top Classifier continuously works until completing all classifications, explaining some of the high peaks in daily activity. Considering the start and end times for each classification, a median classification time of approximately one minute is found (Fig 8C). While this time varies from month to month, it gives evidence of a decrease after March due to improvements in the workflow structure that enhanced efficiency. In particular, 50% of classifications were completed within 30–100 seconds (Fig 8D), 24% took longer than 100 seconds, and 26% took less than 30 seconds, with 22.4% of classifications being completed within 10–30 seconds. Longer classification times may reflect volunteers reading the tutorial during classification or leaving the computer unattended, whereas shorter times often occur when volunteers do not find the required town or encounter an empty line. It is estimated that the minimum time required to scroll through all tasks is about 12 seconds. Finally, it is noteworthy that most classifications (95.7%) completed in 10–20 seconds were performed by the Top Classifier, whose experience allows for significantly faster processing. Interestingly, the data also allow us to infer the UTC time zone of each volunteer (Fig 9A). Their spatial distribution indicates that most volunteers reside in the USA and Europe. Some inaccuracies may exist, as Zooniverse does not account for the transition between summer and winter time. Furthermore, considering only the subset of 188 volunteers (16.3% of the total) who declared their country of origin (Fig 9B), 39% were from the USA, 10% from the UK, and 8%, 7%, and 4% from Italy, Germany, and France, respectively. The remaining volunteers came from 37 other countries. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 9. A) Percentage of volunteers for each UTC band; B) Percentage of volunteers for each country. The information about the borders comes from https://www.naturalearthdata.com/. https://doi.org/10.1371/journal.pclm.0000865.g009

[END]
---
[1] Url: https://journals.plos.org/climate/article?id=10.1371/journal.pclm.0000865

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/