Proceedings of OSM Science 2024

Proceedings of OSM Science 2024 Editors: Marco Minghini A. Yair Grinberger Peter Mooney DOI: 10.5281/zenodo.17370615 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya Contents Editorial of OSM Science 2024 Peter Mooney, A. Yair Grinberger and Marco Minghini 1 Crisis Mapping: Teaching High School ELL Students How To Make Maps That Save Lives Joel Thomas 5 The role of crowd-mapping and PGIS in post-emergency humanitarian operations Valeria Rossi 9 Analyzing the Spatial Distribution of Fuel Stations in Harare, Zimbabwe: Leveraging OpenStreetMap for Disaster Preparedness, Mitigation and Recovery Kingsley Chika Chukwu, Letwin Pondo and Charles Paradzayi 13 Assessing the performance of AI-assisted mapping of building footprints for OSM Anna Zanchetta, Kshitij Sharma and Omran Najjar 17 Shifting trends in global evolution of corporate mapping in OSM Lilly Tockner Benjamin Herfort and Sven Lautenbach 22 Analysis of renewable energy infrastructure representations in OpenStreetMap Luisa Lo Presti and Peter Mooney 26 From Complexity to Clarity: Simplifying OpenStreetMap Data for Improved Active Transportation Analysis Achituv Cohen and Trisalyn Nelson 30 Beyond the seventh mountain, beyond the seventh river – OpenStreetMap as a base map in geographical research Pawel Strus, Anna Górska, Wojciech Brzyski, Aleksandra Bobrek and Natalia Konderak 34 What happens when VGI is threatened? An analysis of the events behind the introduction of rate limiting in OpenStreetMap A. Yair Grinberger and Marco Minghini 34 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya Investigating Corporate Editors in OpenStreetMap Alex Hoferek, Robert Soden and Dipto Sarkar 41 Assessing the attribute accuracy and logical consistency of road data in OpenStreetMap Wangshu Wang and Alexander Zipf 45 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya Editorial of OSM Science 2024 Peter Mooney1,*, A. Yair Grinberger2 and Marco Minghini3,† 1 Department of Computer Science, Maynooth University, Ireland; peter[email protected] 2 Department of Geography, The Hebrew University of Jerusalem, Jerusalem, Israel; yair[email protected] 3 European Commission, Joint Research Centre (JRC), Ispra, Italy; [email protected]opa.eu * Author to whom correspondence should be addressed. † The views expressed are purely those of the author and may not in any circumstances be regarded as stating an official position of the European Commission. In August of this year, 2024, OpenStreetMap (OSM) celebrated its 20th birthday. OSM has grown from a small UK-based mapping project into the largest crowdsourced and volunteered geospatial database in the world. Here, as part of the seventh edition of the Academic Track at the annual State of the Map (SotM) conference, we celebrate the first global SotM on the continent of Africa. This year’s OSM Science, as is the case with all previous editions of the Academic Track, delivers a consolidated knowledge hub for the gathering, sharing, and reporting of scientific progress in OSM-related research. Crucially, this platform facilitates the sharing of scientific findings about OSM directly with the broader OSM community. SotM 2024 takes place from the 6th to 8th September 2024 in Nairobi, Kenya with OSM Science and the Academic Track taking place on Sunday September 8th in its own dedicated session. There are 11 short papers corresponding to 7 full talks and 2 lightning talks presented at the conference with an additional 2 lightning talks presented as video talks. In this Editorial, we summarize these papers by briefly discussing their contributions and attempting to group these works into broad but also interrelated research topics. While this grouping reflects our own interpretation of the contributions and research outcomes of each paper, each extended abstract within the proceedings provides more details and information about the corresponding research. We encourage you to look at these works. Zanchetta et al. [1] consider the challenge of assessing the performance of AI-assisted mapping of building footprints for OSM. Arising from the recent use of AI for mapping assistance in OSM, this work introduces fAIr ( “Free and open source AI for Responsible mapping"): a fully open AI-assisted mapping service to generate semi-automated building footprints features developed by the Humanitarian OpenStreetMap Team (HOT). fAIr displays novelty in reintroducing the “human in the loop” for AI-assisted mapping. Currently, as reported by the authors, fAIr allows OSM mappers to create their own local training dataset, train/fine-tune a pre-trained Eff-UNet model, and then map directly within OSM with the assistance of their own local model. fAIr’s performance is analyzed using twenty five urban regions and seeks to represent different grades of urban characteristics from different regions of the world. Within the domain of assessing the performance of various mapping approaches, Wang and Zipf [2] present their work on assessing the attribute accuracy and logical consistency of road data in OpenStreetMap. Mooney, P., Grinberger, A.Y., & Minghini, M. (2025). Editorial of OSM Science 2024. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.). Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024. Available at https://zenodo.org/communities/osmscience-2024. DOI: 10.5281/zenodo.17333473 © 2025 by the authors. Available under the terms of the Creative Commons Attribution (CC BY 4.0) license. 1 Proceedings of the OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya This work extends the substantial body of existing research work on OSM data quality by exploring quality dimensions specific to the case of motorized vehicle traffic. It does so by analyzing the attribute accuracy, i.e. whether tag values are available and correct, and logical consistency, i.e. topological correctness, of roads while taking into consideration country-specific laws and regulations. The framework is applied for Germany, relating mostly to speed limit data (in the context of attribute accuracy). Cohen and Nelson [3] also consider the transportation domain with their work on a methodological framework which combines OSMnx and a multilane detection algorithm to produce a Simplified OSM Data (SOD) network. The goal of this simplification process is to use the OSM data for improved active transportation analysis. The simplification relates to the combination of multiple transportation lanes related to active transportation (AT) into single centerlines, thus promoting the computation of various measures that relate to self-propelled humanpowered travel modes, e.g. walking and cycling. Ways, in OSM, which share the same name, that show similarity in terms of orientation, and that are in close proximity to each other, are identified as multilane features. The approach is applied to several cities across the world, displaying varying levels of success. Chukwu et al. [4] investigate the use of OSM data for disaster preparedness in Harare, Zimbabwe. Geospatial data about man-made structures and gasoline stations were added to OSM (where they were missing) and this data was then used to perform risk assessment analyses leading to the identification of danger zones and highly vulnerable areas, especially to fire hazards. Connected to disaster preparedness, Thomas [5] reflects on the outcomes and learning experiences of a Mini-Mapathon project, which involved 28 students within three Human Geography courses at an international high school in Beijing, China. Students were taught the basics of humanitarian mapping and then involved in real-world mapping projects using the Humanitarian OpenStreetMap Team’s (HOT) Tasking Manager. A survey allowed the author to assess students’ learnings as well as their motivation and future expectations to remain engaged in mapping projects. We also acknowledge the need to prepare for future climate scenarios and improve sustainability in our interactions with the environment. With this in mind, Lo Presti and Mooney [6] report on an analysis of renewable energy infrastructure representations in OpenStreetMap. To support the renewable energy transition there is a constant need for availability of reliable data about renewable energy infrastructure. In this work, the authors argue that OSM can meet these demands but it is necessary to identify common mapping errors and tagging issues associated with wind and solar energy infrastructures representation within OSM to ensure data quality. The work also presents the results of a geographical analysis to consider the distribution of infrastructures across various land cover types, seeking to detect patterns around land cover and renewable energy infrastructures. Ireland and Belgium are considered in the analysis but extension to more regions and areas is planned in the future. A further example of the use of OSM in the research process is proposed by Strus et al. [7], who involved a group of university students in adding data to OSM through a mix of sources: field mapping using mobile apps, high-resolution imagery, external sources such as scientific publications and a GPS receiver. In addition to providing students with the necessary knowledge to become experienced OSM contributors, the maps generated in this way are then used as a basis for multiple studies in the hydrogeological and urban planning domains. Tockner et al. [8] consider the changing trends in the global evolution of corporate mapping in OSM. In this work, the authors provide an analysis from continuously monitoring 2 Proceedings of the OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya the impact of large corporate contributors to OSM in regards to their changing patterns of contributions. This is very important towards supporting a stronger, data-driven, understanding of the sustainability of OSM in the longer term. Following on from existing research, this work presents analysis to quantify the impact of corporate editing, particularly on volunteer mapping behavior, as an important measure of the structure and robustness of the OSM community. Along the same theme, Hoferek et al. [9] investigates corporate editors in OSM by considering the demographic makeup, careers, motivations and community relationships of the OSM contributors belonging to corporate editors. Results of a survey indicated multiple groups of corporate editors. Although interesting results on the corporate editors’ profiles may be extracted the authors concede that a higher rate of participation from corporate editors in these types of surveys is required to ensure better overall representation. The role of humanitarian projects is often considered in the same conversations as those of corporate editors. In this vein, Rossi [10] describes research on examining how crowd-mapping and PGIS (Participatory GIS) could turn out to be a crucial tool in contributing to humanitarian projects, such as HOTOSM, by playing a role in shaping participatory processes. Fieldwork, reported as part of the research, includes an emergency intervention led by a five NGOs consortium project. The team participated in the HOTOSM Project by mapping different polygons/buildings delimiting territorial areas hit by the earthquake, as a part of an initiative by the Open Mapping Hub for West and Northern Africa. The final paper in our proceedings by Grinberger and Minghini [11] reflects on a systems perspective analysis of the events behind the introduction of rate limiting in OSM. This work uses a specific case in OSM’s history – a set of politically-motivated edits deleting and distorting OSM data in Israel – to discuss the vulnerabilities and resilience-enhancing capabilities of the project. By tracing how the reactions to these events had led to the formation of a temporary coalition between the local mapping community and the data working group that was eventually used to mobilize support for a project-wide change - the introduction of a daily rate limit – it shows that the basic assumptions of OSM simultaneously increases the project’s vulnerability and promotes resilience. Overall, our proceedings reflect a wide range of topics, from AI tool application and data quality to sustainability, education, and humanitarian applications. As with SotM conferences of the past, this year’s conference continues to demonstrate the diverse and impactful research being conducted within the OSM community. It is thus with great excitement that we look to the future of research with and about OSM, hoping that this OSM Science 2024 meeting will continue to provide contributions to shape and explore the field of scientific exploration of OSM and its long-term research agenda. References [1] Zanchetta, A., Sharma, K., & Najjar, O. (2024) Assessing the performance of AI-assisted mapping of building footprints for OSM. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 17–21. [2] Wang, W., & Zipf, A. (2024) Assessing the attribute accuracy and logical consistency of road data in OpenStreetMap. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 45–48. [3] Cohen, A., & Nelson, T. (2024) From Complexity to Clarity: Simplifying OpenStreetMap Data for Improved Active Transportation Analysis. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 30–33. 3 Proceedings of the OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya [4] Chika Chukuw, K., Pondon, L., & Paradzayi, C. (2024) Analyzing the Spatial Distribution of Fuel Stations in Harare, Zimbabwe: Leveraging OpenStreetMap for Disaster Preparedness, Mitigation and Recovery. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 13–16. [5] Thomas, J. (2024) Crisis Mapping: Teaching High School ELL Students How To Make Maps That Save Lives. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 5–8. [6] Lo Presti, L., & Mooney, P. (2024) Analysis of renewable energy infrastructure representations in OpenStreetMap. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 26–29. [7] Strus, P., Górska, A., Brzyski, W., Bobrek, A., & Konderak, N. (2024) Beyond the seventh mountain, beyond the seventh river - OpenStreetMap as a base map in geographical research. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 34–36. [8] Tockner, L., Herfort, B., Lautenbach, S. (2024) Shifting trends in global evolution of corporate mapping in OSM. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 22–25. [9] Hoferek, A., Soden, R., & Sarkar, D. (2024) Investigating Corporate Editors in OpenStreetMap. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 41–44. [10] Rossi, V. (2024) The role of crowd-mapping and PGIS in post-emergency humanitarian operations. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 9–12. [11] Grinberger, A.Y., & Minghini, M. (2024) What happens when VGI is threatened? An analysis of the events behind the introduction of rate limiting in OpenStreetMap. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.) Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024, 37–40. 4 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya Crisis Mapping: Teaching High School ELL Students How To Make Maps That Save Lives Joel Thomas1,* 1 University of Miami, Miami, United States; joelthomasteache[email protected] * Author to whom correspondence should be addressed. This abstract was accepted to the OSMScience 2024 Conference after peer-review. This conference paper presents the learning experiences and outcomes of a Mini-Mapathon course developed, implemented, and evaluated for the first time in a population of 28 Asian Junior high school English Language Learner (ELL) students in three Human Geography courses at an international high school in Beijing, China. The paper includes an introduction to crisis mapping, an educational website built by the teacher, and a description of the project-based learning Mini-Mapathons. Prior to the high school crisis mapping courses, three Mini-Mapathons were piloted for 75 Asian 4th grade students in three classrooms after they had learned about crises in their IB-PYP curriculum. Kilometers of roads and number of buildings mapped were measured collectively and shared with participants after each Mini-Mapathon. The paper concludes with an analysis of students’ knowledge and skill gains, and attitudes towards map making. Other variables measured included students’ interest and motivation for participating in future Mapathons or starting YouthMappers chapters in their future colleges or universities. Students’ survey responses were analyzed using mixed methods. A recommendation for further research is proposed. Never has crisis mapping been more relevant, as disasters are increasing in frequency and severity, resulting in billions of dollars in economic loss and a record-breaking number of people being affected [1, 2]. Crisis mapping is defined as “the real-time gathering, visualizing, and analysis of data during conflict and disaster settings” [3]. In response to disasters, a new form of humanitarianism has emerged: digital humanitarianism in the form of crisis mapping [4]. Crisis mapping is an interactive participatory approach revolutionizing the way crises are understood, reported, and managed. A Mini-Mapathon, a 45-60 minute version of a typical three hour Mapathon, is one way to co-create maps so that vulnerable populations, governments, and NGO’s can respond to crises as well as increase awareness and preparedness to reduce risks when disaster strikes. The Mini-Mapathon project objectives were three-fold. First, to teach students how to save lives through maps. Second, to provide students a real world educational and volunteering experience that meets their service-learning requirements. Third, to enhance students’ Human Geography knowledge and skills. Using the ADDIE (Analyze, Design, Develop, Implement, Evaluate) approach, a Mini-Mapathon curriculum was designed and implemented to answer three research questions below [5]. 1. How did the design of Mini-Mapathons support changes in students’ perceptions of their crisis mapping knowledge and skills? Thomas, J. (2025). Crisis Mapping: Teaching High School ELL Students How to Make Maps That Save Lives. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.). Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024. Available at https://zenodo.org/communities/osmscience-2024. DOI: 10.5281/zenodo.17352871 © 2025 by the authors. Available under the terms of the Creative Commons Attribution (CC BY 4.0) license. 5 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya 2. How did the participation in the Mini-Mapathon support students’ expression of interest in crisis mapping in the future? 3. How did the Mini-Mapathon support students in engaging crisis mapping? Twenty-eight Asian high school students participated in the Mini-Mapathon so that medicine and services could be delivered to vulnerable populations in Nigeria and Zimbabwe by Doctors Without Borders and The Red Cross. The Mini-Mapathons for ELL elementary students and ELL high school students were inspired by the Mini-Mapathons launched effectively with 10 year old students in Milan, Italy [6]. After www.hotosm.org accounts were created and students embarked on a virtual field trip to meet the vulnerable population in their geographic area via 2-3 minute YouTube videos from UNHCR, Red Cross, or Doctors Without Borders, step by step in class instructions were implemented to make learning visible and meet the academic language needs of the ELL students [7,8]. During the 45-minute Mini-Mapathons, students were taught how to map buildings and roads using software on www.hotosm.org. Pre and post surveys were collected so that researchers could better understand and measure how students learned during Mini-Mapathons and determine if this innovative methodology for teaching crisis mapping was effective [5]. Figure 1. Photos from the Mini-Mapathon with ELL students for Chad, Africa In response to RQ 1 How did the design support changes in students’ perceptions of increasing their crisis mapping knowledge and skills? findings indicated that ELL students effectively increased their geospatial knowledge and skills following their experiential learning of map-making because they were able to follow instructions and construct a crisis map [9]. The process of students learning was through the concrete, transformational experience of creating a map that could be used in the real world [9]. Students’ self-assessment of their skills also indicated improvement after the Mini-Mapathon intervention with Pre Survey Q.3. How would you rate digital map making skills? mean value 2.92 compared to a mean value of 3.27 in Post Survey Q.5. How would you rate your digital map making skills?. Many students reported they learned that maps could be collaboratively created by many people and international relief organizations need our help showcasing the tenets of Connectivism Theory at work [10]. Finally, through this innovative learning experience, participants were able to gain service-learning hours to meet their requirement to graduate high school through this project-based learning (PBL) Mini-Mapathon. In response to Post Survey Q.3 What didn’t you know before?, when given multiple choice along with an open ended answer, 43% of the students indicated they didn’t know that maps can be made by many people, 32% learned that maps can be created, and 21% learned 6 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya Analyzing the Spatial Distribution of Fuel Stations in Harare, Zimbabwe: Leveraging OpenStreetMap for Disaster Preparedness, Mitigation and Recovery Kingsley Chika Chukwu1,*, Letwin Pondo2 and Charles Paradzayi3 1 Environmental Information System, University of Lay Adventists of Kigali, Rwanda; [email protected] 2 Surveying and Geomatics Department, Midlands State University, Zimbabwe; [email protected] 3 Faculty of Built Environment, Art and Design, Midlands State University, Gweru, Zimbabwe; par[email protected] * Author to whom correspondence should be addressed. This abstract was accepted to the OSMScience 2024 Conference after peer-review. Natural or man-made disasters pose serious risks to communities all over the world and frequently have dire repercussions, including the loss of life, destruction of property, and disruption of social order [1]. Fire occurrences are particularly dangerous among these calamities, especially when they include infrastructure such as gasoline stations [2]. The potential for large-scale fire catastrophes underscores the need to ensure the safety of gasoline station infrastructure, as evidenced by occurrences documented in Zimbabwe [3]. Petroleum derivatives, such as gasoline, diesel, kerosene, and liquefied petroleum gas (LPG), have the potential to ignite fires when handled improperly [4]. It is clear that a comprehensive fire danger assessment is necessary, which highlights the need for proactive fire management planning [4]. Between 1993 and 2004, there were around 243 fire-related incidents at fuel service stations worldwide that were recorded [5]. It is evident that these sites present significant risks. Geospatial technology has become an invaluable instrument for disaster preparedness, response, and mitigation as a result of these issues [6]. By offering open data necessary for disaster response and mitigation, initiatives such as the Humanitarian OpenStreetMap Team (HOT) and the utilization of platforms similar to OpenStreetMap (OSM) have made substantial progress in these efforts [7]. These systems ensure the availability of high-quality data for efficient mitigation measures and provide quick and open access to geospatial data, facilitating prompt disaster response activities [8]. The OSM data consists of many datasets that use points, lines, polygons, and area attributes to represent real-world features [9,10]. These databases include characteristics that are useful for study, mitigation, recovery, and preparedness for disasters. Through querying the OSM database using the QuickOSM plugin [11] on the Quantum Geographic Information System (QGIS) [12], there were only 54 petrol stations in Harare Chukwu, K. C., Pondo, L., & Paradzayi, C. (2025). Analyzing the Spatial Distribution of Fuel Stations in Harare, Zimbabwe: Leveraging OpenStreetMap for Disaster Preparedness, Mitigation and Recovery. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.). Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024. Available at https://zenodo.org/communities/osmscience-2024. DOI: 10.5281/zenodo.17352656 © 2025 by the authors. Available under the terms of the Creative Commons Attribution (CC BY 4.0) license. 13 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya listed in the OSM database. After identifying this gap, petrol station addresses were collected from the Zimbabwe Petroleum Regulatory Authority (ZPRA). This revealed that there are actually 324 petrol stations in Harare, meaning 288 petrol stations are missing from the OSM database. Using GIS techniques and the Java OpenStreetMap Editor (JOSM) [13] and EveryDoor [14] open-source mobile application, the remaining 288 petrol stations were mapped into the OSM database. This research aimed to address the challenges posed by fire hazards, particularly in petrol station infrastructure, through the application of GIS techniques. The study conducted geoprocessing analyses, including overlay and proximity analysis, using building footprints, conventional construction standards for the siting of petroleum liquid and gas facilities, and public spaces such as markets, parks, schools, hospitals, places of worship, and road networks in Harare, Zimbabwe. Also, nearest neighbor analysis and Euclidean distance analysis [15] were performed using petrol station points to identify danger zones and potential vulnerabilities. The study also sought to map petrol stations, obtain their exact locations, and extract pertinent data from OSM and other sources using GIS methodologies. The study also aims to categorize areas at risk of fire hazards based on various factors, such as proximity to fuel stations and the location of filling stations according to Zimbabwean government regulations and retail premises, as well as other factors identified through spatial analysis. Furthermore, it sought to offer practical suggestions and solutions for improving public safety and lessening the effects of fire disasters in Harare, Zimbabwe, offering insightful information for initiatives related to disaster preparation, mitigation, and recovery to advance resilience and sustainable development in the country while also adding to the scientific understanding of the dangers of fire hazards related to gasoline stations as well as providing a replicable approach to mapping external data into the OSM database. This study's methodology combines spatial analytic methodologies and data extraction from GIS analysis. To fully comprehend the gaps that now exist, the research first starts by gaining access to the data that is already available on the OSM Database. The gasoline stations missing in the OSM database were then mapped into the database, where pertinent data such as locations and infrastructure aspects were recorded. Then relevant data was extracted from OSM. The procedure involved extracting comprehensive information on the locations of buildings, roads, and public spaces, including playgrounds, marketplaces, hospitals, and schools, from the massive OSM database. This extraction was done using the QuickOSM plugin on QGIS software, which shows the richness of the OSM database and the procedures used in this study are provided and detailed in the methodology section. All geospatial datasets, including those from OSM and the Zimbabwe Petroleum Regulatory Authority (ZPRA), are attached and accessible for replication of the analysis. A thorough examination of the geographic distribution of filling stations, public buildings, and fire service stations was done using the GIS technique. The objective of this research was to categorize areas at risk of fire hazards according to the Zimbabwean government's standards for the placement of retail and LPG filling stations. These rules include factors such as the distance from fuel stations and other infrastructural concerns. To fully comprehend the spatial distribution of fuel stations in relation to fire stations, residential buildings, and public amenities like schools, hospitals, places of worship, markets, parks, and more, techniques like Euclidean, nearest neighborhood analysis, and 14 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya geoprocessing analysis were used. The mapping results are presented as maps, graphs and tables, where the identified risk zones are visualized, and detailed analysis is provided. To perform the analysis, the number of buildings in each of the four fire danger risk zones—very high, high, medium, and low risk—was grouped and examined. In order to determine the number of structures in each of these designated zones, a query to the OSM dataset was made. This yields important information on the possible effects of fire hazard events on different regions within the research area. This study aimed to provide crucial information for risk and disaster management (preparedness, mitigation, and recovery) by conducting a thorough investigation of fire hazard threats associated with petrol stations in Harare, Zimbabwe. Utilizing GIS techniques and geospatial data from OSM, the project sought to improve public safety, mitigate the effects of fire disasters, and advance the scientific understanding of fire hazard risks. The methodology, which included collaboration with government agencies, ensured that essential data, such as petrol station addresses and coordinates provided by the Zimbabwe Petroleum Regulatory Authority (ZPRA), were accurately uploaded and integrated into the OSM database. This was achieved using the JOSM, with validation through the EveryDoor mobile application, ensuring data alignment with OSM’s structure and protocols. The study identified significant data gaps within OSM, particularly the absence of many petrol stations in Harare, which were addressed by integrating the ZPRA datasets. While OSM proved to be a valuable source of geospatial data, the study highlights the need for continuous updates and systematic data validation to enhance its utility. Recommendations include encouraging policymakers to leverage OSM for disaster risk reduction strategies and fostering collaborations between government agencies and the OSM community to ensure data accuracy and relevance. The findings and recommendations of this study emphasize the importance of integrating external datasets with open-source geospatial databases to improve hazard mapping and risk management efforts. By doing so, it is possible to enhance public safety and disaster preparedness, making OSM a more reliable tool for both local and global disaster response initiatives. References [1] Kumar, V. (2024). Man-Made Disasters in India. Applied Sciences Research Periodicals, 2(4), 30–42. [2] Kiranmayee, T., Purushotham, B., Vyshnavi, C., Kumar, K. V., Prameeth, M. S., & Kumar, B. V. R. (2024). Fire Prevention in Buses Using Water Head and AWO System. International Journal for Research in Applied Science and Engineering Technology, 12(IV), 1517–1521. [3] Chibwe, J., & Khan, M. A. (2020). A study of adherence to occupational health and safety standards of fuel service stations in Harare. Proceedings of the GARA’s International e-Conference on Enriching e-Learning Management for Global Education: New Norm Viewpoint, 19-20 December 2020, 201–220. [4] Ahmad, M., Widodo, P., Utomo, S. T., & Kusuma, K. (2024). The Efforts of Lombok Regency Government to Recovery After the 2018 Lombok Earthquake Disaster. International Journal of Humanities Education and Social Sciences, 3(5), 2814–2823. [5] Ahmed, M. M., Kutty, S. R. M., Shariff, A. M., & Khamidi, M. F. (2011). Petrol fuel station safety and risk assessment framework. Proceedings of the 2011 National Postgraduate Conference, Perak, Malaysia, 19-20 September 2011, 1–8. [6] Aggarwal, A. (2016). Exposure, hazard and risk mapping during a flood event using open source geospatial technology. Geomatics, Natural Hazards and Risk, 7(4), 1426–1441. 15 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya [7] Herfort, B., Lautenbach, S., Porto de Albuquerque, J., Anderson, J., & Zipf, A. (2021). The evolution of humanitarian mapping within the OpenStreetMap community. Scientific reports, 11(1), 3037. [8] Sunday, V. N., Ndukwu, R. I., & Brovelli, M. A. (2022). Analysis of local and remote mappers’ open geographic data contribution to oil spill disaster response in Niger delta region, Nigeria. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-4/W1-2022, 465–472. [9] Minghini, M., Gonzalez, S. T., & Gabrielli, L. (2024). Pan-European open building footprints: analysis and comparison in selected countries. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-4/W12-2024, 97–103. [10] Hacar, M., & Gökgöz, T. (2017). Determining Coordinate Transformation Method for the Preprocess of Road Matching. International Symposium on GIS Applications in Geography & Geoscience. [11] Trimaille, E. (2024). QuickOSM — QGIS Python Plugins Repository. Retrieved from https://plugins.qgis.org/plugins/QuickOSM/#plugin-details [12] QGIS.org (2024). QGIS. Retrieved from http://qgis.org [13] JOSM (2024). Java OpenStreetMap Editor. Retrieved from https://josm.openstreetmap.de [14] Zverev, I. (2024). Every Door. Retrieved from https://every-door.app [15] Gul, S., Ali, Z., & Ullah, U. (2024). Integrating GIS and Remote Sensing for Comprehensive Flood Risk Zonation in Tehsil Shah Alam (Peshawar). Journal of Asian Development Studies, 13(2), 142– 1431. 16 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya Assessing the performance of AI-assisted mapping of building footprints for OSM Anna Zanchetta1,*, Kshitij Sharma2 and Omran Najjar2 1 Alan Turing Institute, London, United Kingdom; [email protected] 2 Humanitarian OpenStreetMap Team; [email protected]g, [email protected] * Author to whom correspondence should be addressed. This abstract was accepted to the OSMScience 2024 Conference after peer-review. Building footprints features are useful in a wide range of applications such as disaster assessment, urban planning, and environmental monitoring [1-3], and their identification has been gaining increasing interest and attention from the machine learning (ML) research in Earth Observation [4]. In the context of disaster response, accurate and prompt availability of such information is crucial [5-7]. Open source datasets of AI-generated building footprints exist (e.g. Microsoft’s global buildings dataset, available through Rapid, and Google’s Open Buildings for Africa and the Global South at large); however, the ML models underlying these datasets are not currently open sourced [8]. fAIr, a fully open AI-assisted mapping service to generate semi-automated building footprints features developed by the Humanitarian OpenStreetMap Team (HOT) [9], addresses this need. fAIr stands for “Free and open source AI for Responsible mapping, that is resilient to local contexts and relates to local communities", reflecting the objective of HOT to improve and assist mapping for humanitarian aid and disaster relief. In particular, fAIr performs semantic segmentation to detect building footprints from imagery at high resolution (cm) openly available through OpenAerialMap (OAM) [10]. OAM is an open service that provides search and access to openly licensed satellite and unmanned aerial vehicle (UAV) imagery uploaded by users to its website. While in OSM building footprints mapping is currently supported in most countries through Rapid, the need remains for adjustments and corrections [11]. fAIr addresses this issue by reintroducing the human in the loop. In its current state, fAIr allows OSM mappers to create their own local training dataset, train/fine-tune a pre-trained Eff-UNet model, and then map into OSM with the assistance of their own local model. In its initial release [8], the performance of the model following training was not assessed; the objective of this research is to address this gap. This paper describes the research developed to assess how the ML fine-tuning process performs, investigating the currently used accuracy metric, and comparing against different sets of evaluation metrics. This falls within the broader spectrum of research on understanding the fine-tuning process for geographic domain adaptation in image analysis validation [12,13]. The ultimate aim of this research is to advise on the performance of fAIr, by assessing the current validation accuracy and comparing against other accuracy metrics for building footprints segmentation tasks. Zanchetta, A., Sharma, K., & Najjar, O. (2025). Assessing the performance of AI-assisted mapping of building footprints for OSM. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.). Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024. Available at https://zenodo.org/communities/osmscience-2024. DOI: 10.5281/zenodo.17352770 © 2025 by the authors. Available under the terms of the Creative Commons Attribution (CC BY 4.0) license. 17 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya To test fAIr’s performance, we selected twenty five urban regions from OAM imagery, seeking to represent different grades of urban characteristics from different regions of the world. The urban regions were categorised by three variables through visual inspection: urbanity, meaning the degree of urbanity per each region, with classes: ‘rural’, ‘peri-urban’, ‘urban’, ‘refugee camp’; urban density, with classes: ‘sparse’, ‘dense’, ‘grid’ (a class in between sparse and dense, which shows a regular gridded or repeated spatial pattern); roof cover type, with classes: ‘shingles’, ‘metal’ (usually tin), ‘cement’, ‘mixed’ (a mix of the above, or other types). We then performed manual labelling on selected areas of interest (AoI), leveraging OSM data when this was available and aligned, or else generating new OSM data while labelling the images (see fAIr-dev website [14] for a detailed explanation on the labelling process). The pre-processing of the AoI images, to generate one training dataset per urban region, was performed through fAIr-dev website [14] and produced 256x256 georeferenced tiles for both the original RGB images (ground truth) and OSM data (the labels), where the latest were converted to binary masks for the analysis. The number of tiles per region varies depending on the training dataset, for a total of 8400 images (about 350 per region in average) at three different zoom levels (19, 20, 21). Having labelled the AoIs, the next step was semantic segmentation. In computer vision (CV), semantic segmentation is the task of segmenting an image into semantic meaningful classes, which is performed with convolutional neural network (CNN) architectures [15]. The deep learning CNN model used in fAIr is called RAMP (Replicable AI for MicroPlanning), and its architecture originates from a typical encoder-decoder network, a computational process where images are trained on labelled data, called Eff-UNet model [16]. This consists of an EfficientNet as the encoder for feature extraction, with UNet decoder for reconstructing the segmentation map [17]. RAMP was chosen as an outcome of a series of feasibility studies and AI challenges to segmenting buildings for disaster resilience, happening between 2020 - 2022 and supported by HOT and other international bodies [8]. fAIr is trained using categorical accuracy (i.e. the ratio of samples that were correctly classified over all predictions made) at the pixel level as an accuracy metric, with categorical cross entropy as the loss function. Based on a literature review of the performance of metrics in image analysis validation, we chose to compare the current metric against four commonly used validation metrics [13; 18]: precision, also known as Positive Predictive Value (PPV), the fraction of actual positive samples among the positive predictions; recall, also called sensitivity or True Positive Rate (TPR), the fraction of true positive predictions over all positive samples; F1 score, the harmonic mean of PPV and TPR; intersection over union (IoU), also known as Jaccard index, a measure of the overlap between the predicted segmentation and the ground truth. We ran the ML training on all twenty five training datasets using an Nvidia Tesla T4 GPU, at four different batch sizes (2, 4, 8 and 16), and for a number of epochs fixed to 20, to be consistent with the current functioning and suggestions of fAIr. For each urban region, the training provided a fine-tuned model backbone, and the value at each epoch of the five metrics, defined using built-in Tensorflow functions. In order to use the fine-tuned model backbones for inference later on, fAIr needs to store the model checkpoints. This is currently done using early-stopping, which means that the model is stored at the epoch at which the categorical accuracy has its maximum value for validation during training. In our analysis we consider this epoch, per each urban region, as the benchmark against which to evaluate the 18 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya model performance. In other words, when categorical accuracy is at its highest, how do other metrics perform? In particular, how about IoU, which is the suggested metric for this type of image analysis problem [13]? In a preliminary analysis we compared the values of each metric against one another, at the epoch at which categorical accuracy was the highest for validation after training. This was done taking into account the three variables mentioned above, and it showed that the degree of urbanity and roof type do not seem to have a relevant influence on the performance, at least for the twenty five regions considered in the analysis, and were therefore discarded. When considering only the density variable, an analysis on the batch size showed that the distribution of the metrics is consistent across the batch sizes, with variations being attributable mainly to density; therefore the batch size is also discarded as a variable. The values for all the metrics grouped by the density classes are plotted in Figure 1 for the four batch sizes (2, 4, 8 and 16), and for all the urban regions. Clearly a trend is seen going from sparse to dense urban regions, with IoU and recall performances being mostly affected. The results thus point to density being the major driver in affecting the model accuracy during validation. Among all the twenty five selected urban regions and for all the accuracy metrics, dense regions perform worse on average, 4.2% less than sparse and 4.8% less than grid regions. These figures grow respectively to 7.8% and 7.7% when considering IoU alone. While ideally both precision and recall should have high values (closer to 1), in our Figure 1. Boxplot with values of five validation accuracy metrics (categorical accuracy, precision, recall, F1 score, IoU) for each urban region (left y axis) and grouped by the density variable (right y axis), for four batch sizes (2, 4, 8, 16, bottom x axis). 19 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya case high values for precision mean that most detected objects match ground truth objects, but low recall implies that many ground truth objects have been missed, and this also affects F1 score. Low values for IoU imply that the model struggles in distinguishing objects from their backgrounds, a behavior that resonates with more densely populated areas in this case. In conclusion, fAIr shows to have more potential for mapping buildings in sparsely populated areas than in urban centres. When using fAIr at its current state, one should be aware that building footprints segmentation in dense areas will be less performant than in gridded and sparse areas, with IoU, the recommended metric for this CV task, giving the lowest model performance. Other factors, like the roof cover type and regionality, do not seem to have a relevant effect on the model performance. Future research will concentrate on using other factors to help drive the fine-tuning process, like loss regularisation, and on how other variables, like the zoom level, can affect the performance. The analysis will also be extended to the assessment of the inference process. All the code used in the research is available at https://github.com/ciupava/fAIr-utilities, while the preprocessed data is downloadable directly from the fAIr-dev website [14]. References [1] You, Y., Wang, S., Ma, Y., Chen, G., Wang, B., Shen, M., & Liu, W. (2018). Building detection from VHR remote sensing imagery based on the morphological building index. Remote Sensing, 10(8). [2] Owusu, M., Kuffer, M., Belgiu, M., Grippa, T., Lennert, M., Georganos, S., & Vanhuysse, S. (2021). Towards user-driven earth observation-based slum mapping. Computers, Environment and Urban Systems, 89, 101681. [3] Yang, J., Matsushita, B., & Zhang, H. (2023). Improving building rooftop segmentation accuracy through the optimization of unet basic elements and image foreground-background balance. ISPRS Journal of Photogrammetry and Remote Sensing, 201, 123–137. [4] Hoeser, T., Bachofer, F., & Kuenzer, C. (2020). Object detection and image segmentation with deep learning on earth observation data: A review—Part II: Applications. Remote Sensing, 12(18). [5] Boccardo, P., & Giulio Tonolo, F. (2015). Remote sensing role in emergency mapping for disaster response. In: Lollino, G., Manconi, A., Guzzetti, F., Culshaw, M., Bobrowsky, P., & Luino, F. (Eds.), Engineering geology for society and territory - Volume 5, Springer, Cham, 17–24. [6] Deng, W. Y., L. (2022). Post-disaster building damage assessment based on improved u-net. Nature Scientific Reports, 12, 15862. [7] Sun, Z., Zhang, Z., Chen, M., Qian, Z., Cao, M., & Wen, Y. (2022). Improving the performance of automated rooftop extraction through geospatial stratified and optimized sampling. Remote Sensing, 14(19). [8] Humanitarian OpenStreetMap Team (2022). hot_tech_talk | fAIr: AI-assisted mapping. Retrieved from https://www.hotosm.org/tech-blog/hot-tech-talks-fair [9] Humanitarian OpenStreetMap Team (2022). fAIr. Retrieved from https://www.hotosm.org/techsuite/fair [10] OpenAerialMap (2017). OpenAerialMap. Retrieved from https://openaerialmap.org [11] OpenStreetMap Wiki (2024). Rapid. Retrieved from https://wiki.openstreetmap.org/wiki/Rapid [12] Rainio, O., Teuho, J., & Klén, R. (2024). Evaluation metrics and statistical tests for machine learning. Nature Scientific Reports, 14(1), 6086. [13] Maier-Hein, L., Reinke, A., Godau, P., Tizabi, M. D., Buettner, F., Christodoulou, E., Glocker, B., Isensee, F., Kleesiek, J., Kozubek, M., et al. (2024). Metrics reloaded: recommendations for image analysis validation. Nature Methods, 21(2), 195–212. [14] Humanitarian OpenStreetMap Team (2024). fAIr development environment. Retrieved from https://fair-dev.hotosm.org 20 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya [15] Hoeser, T., & Kuenzer, C. (2020). Object detection and image segmentation with deep learning on earth observation data: A review—Part I: Evolution and recent trends. Remote Sensing, 12(10). [16] Replicable AI for Microplanning (RAMP) (2020) RAMP model card. Retrieved from https://rampml. global/ramp-model-card [17] Baheti, B., Innani, S., Gajre, S., & Talbar, S. (2020). Eff-unet: A novel architecture for semantic segmentation in unstructured environment. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops, Seattle, USA, 13-19 June 2020, 1473–1481. [18] Reinke, A., Tizabi, M. D., Baumgartner, M., Eisenmann, M., Heckmann-Nötzel, D., Kavur, A. E., Rädsch, T., Sudre, C. H., Acion, L., Antonelli, M., et al. (2024). Understanding metric-related pitfalls in image analysis validation. Nature Methods, 21(2), 182–194. 21 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya Shifting trends in global evolution of corporate mapping in OSM Lilly Tockner1, Benjamin Herfort1,2,* and Sven Lautenbach1,2 1 Heidelberg Institute for Geoinformation Technology, Heidelberg, Germany; lilly[email protected], benjamin.herfor[email protected], [email protected]g 2 GIScience Chair, Institute of Geography, Heidelberg University, Heidelberg, Germany * Author to whom correspondence should be addressed. This abstract was accepted to the OSMScience 2024 Conference after peer-review. In recent years, with the emergence of corporate editing in OpenStreetMap (OSM), there has been interest, and in some cases concern, about its influence in OSM. For instance, large corporations such as Apple, Microsoft, Meta and Amazon have hired large teams to edit in OSM [1]. The launch of the Overture Maps Foundation, by Amazon, Meta, Microsoft and TomTom hosted by the Linux Foundation in 2022 and the release of its first dataset [2] has also led to heated debates on the OSM forum, regarding the future of OSM. Concerns have been raised about the monopolization of geodata, the replacement of OSM by other sites or the backlining of OSM [3]. Consequently, analyzing and continuously monitoring the impact of corporate contributors and continuing to observe their fluctuating patterns of editing will be significant to the understanding of the sustainability of OSM. Therefore, this talk will look at corporate editing in OSM at three scales - global, national and local to answer the two research questions: ● (RQ1) What is the impact of corporate mapping on global scale mapping? ● (RQ2) What is the impact of corporate mapping on country and small-scale mapping? There are two main avenues to track corporate contributors: Either through corporate affiliated OSM User IDs (UIDs) or through corporate hashtags in OSM changesets. UIDs of corporation affiliated mappers have been used in the past to track corporate activity in OSM (e.g. [1,4,5]). A list of affiliated usernames can be collected from disclosed lists on the OSM Wiki or corporations GitHub pages [6]. The second way to track corporate edits is through corporate hashtags in changesets. Since 2009, it became possible to add hashtags to changesets in OSM, and this has become common practice, especially to show affiliation to regions, events, organizations or corporations [7]. For the analysis presented here, in total 24 companies were tracked including Apple, Kaart, Amazon, Microsoft, TomTom, Grab, DigitalEgypt, Mapbox and Meta. We used tracking with hashtags as proposed by [7]. The dataset used for the analysis integrated data from two sources: The OpenStreetMap History Database [8] and the OSM Changeset database. The OSHDB was used to derive OSM geometries and attribute information as well as information about the editors. The data was then intersected with a dataset of country boundaries to assign a mapping to a specific country. The OSM changeset dataset was joined to this contribution dataset. The dataset included a plethora of information, but the most relevant columns for the analysis were the OSM ID, Changeset timestamp, hashtags and User IDs, country, year and month as well as Tockner, L., Herfort, B., & Lautenbach, S. (2025). Shifting trends in global evolution of corporate mapping in OSM. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.). Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024. Available at https://zenodo.org/communities/osmscience-2024. DOI: 10.5281/zenodo.17369178 © 2025 by the authors. Available under the terms of the Creative Commons Attribution (CC BY 4.0) license. 22 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya Norway. We did not process the OSM for larger countries but checked the OHSOME dashboard [5] for more insights. According to this dashboard, since July 2021 wind generator mapping in Ireland and Belgium has risen by 36.5% and 14.3% respectively, while solar generator objects have grown by 1154% and 448%. Significant growing trends are present also in other European countries for the same period. Finally, further research is needed to assess the completeness of other infrastructure-related tags, such as operator, installation date, and generator output values. Figure 1. Example of wind farms near peat bogs in Ireland. Red lines mark 1000-meter buffers around wind farms, with land cover shown according to CLC. The map in the bottom right corner highlights the zoomed area with a blue circle. Acknowledgements This publication has emanated from research conducted with the financial support of Science Foundation Ireland under Grant number 18/CRT/6049. For the purpose of Open Access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. References [1] Cader, C., Pelz, S., Radu, A., & Blechinger, P. (2018). Overcoming data scarcity for energy access planning with open data–the example of Tanzania. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLII-4/W8, 23–26. [2] Alhamwi, A., Medjroubi, W., Vogt, T., & Agert, C. (2017). OpenStreetMap data in modelling the urban energy infrastructure: a first assessment and analysis. Energy Procedia, 142, 1968–1976. [3] Lo Presti, L. (2024) Analysis of Renewable Energy Infrastructure Representations in OpenStreetMap. Retrieved from http://github.com/luisalopresti/OSMRenewables [4] Sustainable Energy Authority of Ireland (SEAI) (2021). Community Energy Resource Toolkit. Retrieved from https://tinyurl.com/54pkfcfy [5] HeiGIT (2024) Ohsome Dashboard. Retrieved from https://dashboard. ohsome.org 29 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya From Complexity to Clarity: Simplifying OpenStreetMap Data for Improved Active Transportation Analysis Achituv Cohen1,* and Trisalyn Nelson2 1 Department of Civil Engineering, Ariel University, Ariel, Israel; [email protected] 2 Department of Geography, UCSB, Santa Barbara, Ca, USA; [email protected] * Author to whom correspondence should be addressed. This abstract was accepted to the OSMScience 2024 Conference after peer-review. Active Transportation (AT) refers to self-propelled, human-powered travel modes like walking and cycling [1]. To effectively analyze AT in urban environments, the availability of accurate street network data is paramount. Detailed information on street layout, connectivity, and accessibility [2] is crucial for supporting the evaluation and understanding of walking and bicycling patterns. Combining this information with demographic and environmental factors such as population demographics, AT-related crashes, and land use not only informs infrastructure improvements but also guides strategy development [3]. Such strategies aim to enhance the safety, efficiency, and attractiveness of AT [4,5], thereby mitigating issues such as traffic congestion, obesity, and air pollution [6]. Street network data comes from various sources, each presenting challenges regarding accuracy and update frequency. OpenStreetMap (OSM) offers a comprehensive, open-source geospatial database that is continuously updated by a global community of contributors [7]. OSM's strengths lie in its detailed coverage of road networks, including pedestrian pathways and bicycle lanes, which are crucial for AT analysis [8]. However, utilizing OSM data for AT analysis involves several challenges. Firstly, the granularity of its network can complicate modeling AT users' movements at the street level. For instance, when assessing AT safety or developing a walkability index, the current data representativeness on OSM requires significant manipulation to mitigate bias. Furthermore, due to OSM's open-editing model, the standards for mapping elements are not consistently defined, leading to contentions and variations in data quality [9]. Contributors often map elements based on personal needs and knowledge, introducing inconsistencies in the dataset [10,11]. As a result, in some areas, all designated lanes for various users are meticulously mapped, while in others, only a single lane representing the presence of a street is depicted [12,13]. Additionally, there are instances where only lanes for motor vehicles are detailed, with scant attention paid to lanes catering to other user groups. These inconsistencies necessitate careful consideration and significant data processing when using OSM for AT analysis to ensure accuracy and reliability. In this study, we propose an innovative solution to generate an axial network specifically designed for monitoring and analyzing AT users, using exclusively OSM data Cohen, A., & Nelson, T. (2025). From Complexity to Clarity: Simplifying OpenStreetMap Data for Improved Active Transportation Analysis. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.). Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024. Available at https://zenodo.org/communities/osmscience-2024. DOI: 10.5281/zenodo.17369360 © 2025 by the authors. Available under the terms of the Creative Commons Attribution (CC BY 4.0) license. 30 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya (https://github.com/achic19/SOD). Our approach simplifies the network while preserving its topology. We apply this solution across diverse spatial contexts, from straightforward geographic regions to complex urban environments. Furthermore, we implement our methodology in several cities worldwide — Turin (Italy), Tel Aviv (Israel), and San Francisco (United State) — each characterized by unique urban structures. The preliminary tasks use OSMnx [4] to acquire OSM street network data, converting it into a graph while correcting topological errors. Then, the data is stored in a geodata table, including polyline geometry, names, and road types. Our algorithm filters out unsuitable roads, like motorways and trunk roads, and replaces roundabouts with their central points. The network is then ready for the multilane detection algorithm. Polylines identified in a multilane scenario are aggregated into a centerline and added to the Simplified OSM Data (SOD) network. Other polylines are added to the SOD network, retaining their original geometry. A multilane scenario is identified when polylines share the same street name, have similar angles, and are close together. Polylines are grouped by street name, and azimuth (0°-180°) narrows the angle range to ensure that parallel lines are considered parallel, regardless of orientation. The Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clusters similarly angled polylines using a 10° radius and a minimum of two samples. Through sensitivity tests, we determined that a 10° threshold effectively includes nearly parallel polylines while excluding those that are significantly misaligned. Outliers with significantly different angles are excluded. The remaining clusters apply a right and left shifted buffer to each polyline. If two or more polylines in the same class overlap by at least 10% in the shifted buffer, they're classified as multilane, and the entire class is replaced with one or more centerlines. The 10% threshold balances sensitivity (detecting true parallel lines) and specificity by minimizing false positives from minor geometric variations, such as slight bends or irregularities. The core idea behind creating a new centerline is to identify its start and end vertices and then add intermediate vertices to preserve the overall shape of multilane polylines. Overlapping buffers are merged into a single polygon, and the two polygon vertices that best define the original polylines are used to establish the new centerline's start and end. Intermediate vertices are then added at regular intervals to maintain the original polylines' orientation. The simplification process introduces significant changes to the locations of many polylines, resulting in issues of continuity, topology, accuracy, and inconsistency in the SOD network. Most problems can be resolved using existing methods, but connecting roundabouts requires extra effort. This process has three steps: transforming the roundabout representation to a point, connecting nearby dead-end polylines, and linking polylines in proximity. Following these steps ensures proper continuity and maintains the network's topology. The case study centered on Turin, Italy, simplified its street network to 10,535 polylines from an initial 55,104. This effort reduced the complexity of 459 streets, converting 155 roundabouts to single points. Despite some challenges near intersections, the final version preserved topology (see Figure 1). 31 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya Figure 1. displays various urban configurations before (represented by the red network) and after (illustrated by the blue network) the simplification process. These configurations range from simpler layouts (Figure 1a to Figure 1c), where streets are represented with a single lane, to intermediate complexity (Figure 1d - Figure 1f) featuring streets with two lanes, and finally, more intricate arrangements (Figure 1g - Figure 1i). For validation, 46 streets in Turin were reviewed, revealing a 63% success rate, 22% with minor flaws, and 9% with partial success. Issues such as threshold errors, external mapping inaccuracies, and ambiguous street configurations affected the results. Similar challenges were faced in Tel Aviv, Israel, where 82% of 22 test streets matched perfectly with the reference network. Yet, 9% had minor issues, and 5% achieved partial identification/success. In San Francisco, the results showed 85% accuracy with 20 streets evaluated. While most streets were correctly simplified, minor flaws were present in a few cases. Overall, the methodology successfully streamlined street networks, although it achieved varying rates of success across different cities. Our study offers a robust methodology for generating axial networks for AT users using OSM data, balancing data simplification with topology preservation. Its successful application across diverse cities such as Turin, Tel Aviv, and San Francisco demonstrate its versatility and effectiveness in varied urban environments. The approach simplifies the complex urban network data, streamlining pedestrian and bicycle analysis while retaining essential details. This enables urban planners and policymakers to better monitor and understand AT patterns, leading to infrastructure improvements that support safer, more efficient, and environmentally friendly urban mobility. 32 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya Our research offers both scientific contributions and practical benefits. Scientifically, it has been applied to evaluate the built environment for pedestrian walkability using a spatial data clustering approach. Additionally, it has been utilized in research to evaluate safety and bike network connectivity, aiming to improve bicycle and pedestrian infrastructure in California cities. Practically, this work has been published on GitHub, making the code accessible to everyone for their specific needs and goals. References [1] Alessio, H. M., Bassett, D. R., Bopp, M. J., Parr, B. B., Patch, G. S., Rankin, J. W., Rojas-Rueda, D., Roti, M. W. & Wojcik, J. R. (2021). Climate change, air pollution, and physical inactivity: is active transportation part of the solution? Medicine and science in Sports and Exercise, 53(6), 1170–1178. [2] Soczówka, P., Żochowska, R., & Karoń, G. (2020). Method of the Analysis of the Connectivity of Road and Street Network in Terms of Division of the City Area. Computation, 8(2), 54. [3] Osama, A., & Sayed, T. (2019). A Novel Approach for Identifying, Diagnosing, and Treating Active Transportation Safety Issues. Transportation Research Record, 2673(11), 813–823. [4] Yi, P., Zhen, H., & Zhang, Y. (2004). Assessment of Traffic Volume, Based on Travel Time, to Enhance Urban Network Operation. Transportation Research Record, 1878(1), 164–170. [5] Muriel-Villegas, J. E., Alvarez-Uribe, K. C., Patiño-Rodríguez, C. E., & Villegas, J. G. (2016). Analysis of transportation networks subject to natural hazards – Insights from a Colombian case. Reliability Engineering & System Safety, 152, 151–165. [6] Nelson, T., Ferster, C., Laberee, K., Fuller, D., & Winters, M. (2021). Crowdsourced data for bicycling research and practice. Transport Reviews, 41(1), 97–114. [7] Bennett, J. (2010). OpenStreetMap, Packt Publishing Ltd, Birmingham. [8] Booth, L., Norman, R., & Pettigrew, S. (2019). The potential implications of autonomous vehicles for active transport. Journal of Transport & Health, 15, 100623. [9] Sehra, S. S., Singh, J., Rai, H. S., & Anand, S. S. (2020). Extending Processing Toolbox for assessing the logical consistency of OpenStreetMap data. Transactions in GIS, 24(1), 44–71. [10] Hochmair, H. H., Zielstra, D., & Neis, P. (2015). Assessing the Completeness of Bicycle Trail and Lane Features in OpenStreetMap for the United States. Transactions in GIS, 19(1), 63–81. [11] Ferster, C., Fischer, J., Manaugh, K., Nelson, T., & Winters, M. (2020). Using OpenStreetMap to inventory bicycle infrastructure: A comparison with open data from cities. International Journal of Sustainable Transportation, 14(1), 64–73. [12] Joshi, A., & James, M. R. (2015). Generation of accurate lane-level maps from coarse prior maps and lidar. IEEE Intelligent Transportation Systems Magazine, 7(1), 19–29. [13] Vitalis, S., Labetski, A., Ledoux, H., & Stoter, J. (2022). From road centrelines to carriageways—A reconstruction algorithm. PLoS One, 17(2), E0262801. 33 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya Beyond the seventh mountain, beyond the seventh river – OpenStreetMap as a base map in geographical research Pawel Strus1,*, Anna Górska1, Wojciech Brzyski1, Aleksandra Bobrek1 and Natalia Konderak1 1 Department of Law, Economy and Administration/Geoinformation Research Team, The University of National Education Commision, Krakow, Poland; [email protected], [email protected]akow.pl, [email protected], aleksandr[email protected]akow.pl, natalia.konder[email protected]akow.pl * Author to whom correspondence should be addressed. This abstract was accepted to the OSMScience 2024 Conference after peer-review. The speech is the result of exercises conducted by employees of the Geoinformation Research Team and students of the UKEN University in Krakow, Poland. The basic assumption we made is that OpenStreetMap can be sufficient as a data provider for various geographical works – as a base map for field exercises, as an almost complete environmental database for some compact area (such as an island or a national park). After reviewing the list of tags describing the geographic environment in OpenStreetMap (OSM) [1], we knew this would be possible. We have selected several research polygons, which we call cartographic polygons. These include the Peljesac Peninsula [2,3] in Croatia (see Figure 1), the Aegean Coast near Thessaloniki in Greece, the area around Lake Inari, the Lemenjoki National Park in Finland, and the wild Bieszczady Mountains in Poland. We selected the training fields so that they were either places with nature close to natural conditions and places significantly transformed by human activity. Usually, these were also important places for some key reasons - for example, on the Peljesac Peninsula, a bridge was built to facilitate communication between the two parts of Dalmatia. Not all places were visited, but we have collected cartographic material for all of them. Before each trip, we trained a group of participants on how to use and supplement OSM. Each participant set up their own user account. Field work consisted of completing the content of OSM as accurately as possible - groups of two people were sent into the field and, using the OsmAnd or EveryDoor applications, they inserted all interesting objects on the map [4]. The rest is small-scale work verification and editing of the map in the JOSM editor. Each stage of work was also preceded by a thorough analysis of official OSM tags [1] – which constitute information about all elements of the natural environment. It was found that the best represented features were those related to relief, land cover and hydrology. In particular, the content regarding land cover (down to a single tree and bush), and the richness of descriptions of relief forms (OSM WIKI, Glossary of landforms [5]) allow the creation of appropriate thematic maps – land cover maps and geomorphological maps. Strus, P., Górska, A., Brzyski, W., Bobrek, A., & Konderak, N. (2025). Beyond the seventh mountain, beyond the seventh river – OpenStreetMap as a base map in geographical research. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.). Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024. Available at https://zenodo.org/communities/osmscience-2024. DOI: 10.5281/zenodo.17369393 © 2025 by the authors. Available under the terms of the Creative Commons Attribution (CC BY 4.0) license. 34 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya The principle adopted in JOSM is that we complete the map to the highest degree of accuracy possible using available data and processing capacity. For example, we supplement the terrain coverage for Poland from the available official orthophoto map from the national geoportal with a terrain resolution of 5 cm. The distinction between land cover types is made by students of higher years of geography, so there should not be too many interpretation errors. Figure 1. UKEN GIS Lab Team and students while listening to the advanced OSM lecture. What information is completed on the map? As for the relief of the land - valleys and valley types, rock walls, erosion undercuts, landslides and landslide niches. An additional module of our work is urban micromapping. We check how accurately we can supplement field data so that they can serve two purposes – for students and spatial planning specialists in the analysis and inventory of urban space, and for people with disabilities as a base for accessibility maps used in applications, e.g. blind. For this purpose, we carried out tests of terrain mapping using a geodetic GPS receiver (STONEX 900A). We chose the area on the campus of our university due to the presence of the remains of an old waterbed supplying water to mills and a city moat – a lot of unevenness, steps, suddenly ending sidewalks, etc. Additionally, we have also started work on old housing estates in Krakow's Nowa Huta district – inhabited mainly by older people, and therefore often beneficiaries of all programs regarding the availability of public facilities and apartments. Approximately one thousand points were measured in the above locations with an accuracy ranging from 2 mm to 1 cm. In further stages, they will serve as the basis for the point cloud made during unmanned aerial vehicle raids. A number of additional works were also carried out as part of the project, e.g. wild waste dumps in the Ojców National Park were mapped and marked (we will most likely use the tag amenity=waste_dump_site). 35 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya All work on the project resulted not only in a significant improvement in the quality of the OSM map, but also in the training of a large team who, in their free time, complete OSM data in their area [6]. Both the map analytical and field stages of the work are still in progress. This is our beginning in the work of OSM, but we would like to use our previous experience in completing OSM data in various places around the world in many task programs, e.g. for UN Mappers and the Humanitarian OpenStreetMap Team (HOT). References [1] Ciepluch, B., Mooney, P., Jacob, R., & Winstanley, A. C. (2009). Using openstreetmap to deliver location-based environmental information in Ireland. SIGSPATIAL Special, 1(3), 17–22. [2] Bartczak, W. (2023). Laboratorium Regionów – Warmia i Mazury (popularyzacja OpenStreetMap). Retrieved from https://openstreetmap.org.pl/2023/laboratorium-regionow-warmia-i-mazury-popularyz acja-openstreetmap [3] Bartczak, W. (2023). OSMP w Chorwacji. Retrieved from https://openstreetmap.org.pl/2022/osmpw-chorwacji [4] Neis, P., & Zielstra, D. (2014). Recent developments and future trends in volunteered geographic information research: The case of OpenStreetMap. Future internet, 6(1), 76–106. [5] OpenStreetMap Wiki (2024). Glossary of landforms. Retrieved from https://wiki.openstreetmap. org/wiki/Glossary_of_landforms [6] Gröchenig, S., Brunauer, R., & Rehrl, K. (2014). Digging into the history of VGI data-sets: results from a worldwide study on OpenStreetMap mapping activity. Journal of Location Based Services, 8(3), 198–210. 36 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya What happens when VGI is threatened? An analysis of the events behind the introduction of rate limiting in OpenStreetMap A. Yair Grinberger1,* and Marco Minghini2,† 1 Department of Geography, the Hebrew University of Jerusaelm, Jerusalem, Israel; yair[email protected] 2 European Commission, Joint Research Centre (JRC), Ispra, Italy; [email protected]opa.eu * Author to whom correspondence should be addressed. † The views expressed are purely those of the author and may not in any circumstances be regarded as stating an official position of the European Commission. This abstract was accepted to the OSMScience 2024 Conference after peer-review. OpenStreetMap (OSM) operates on optimistic principles: it assumes that a diverse crowd can collaboratively agree on project goals and the methods to achieve them. This presumption of goodwill among participants is, for the most part, validated by OSM's success. However, there are instances where these assumptions are challenged, such as disagreements over tag usage [1], the mapping of general feature types [2], or specific entities [3]. When unresolved, these disputes can escalate into "editing wars," where users engage in repeated edits to enforce their views [4]. Despite such conflicts, the involved parties generally do not contest OSM's overarching objective of producing an accurate global map. Deviations from this goal, classified as platform abuse [5], include actions like spamming (e.g., using OSM for promotional purposes) and vandalism, where edits intentionally misrepresent on-the-ground reality. These abuses, particularly vandalism, pose significant risks to the project's success, as an inaccurate map can render OSM unusable in many contexts. Such incidents are not uncommon, as evidenced by an analysis of user bans [6]. Academic studies on this issue have primarily focused on detecting vandalism, e.g. [7], and analyzing specific cases' impacts on data integrity, e.g. [8]. OSM's primary responses to vandalism include post-hoc data reverts, which any user can perform, and user bans issued by the Data Working Group (DWG), the OSM Foundation body responsible for handling data-related issues such as copyright infringement and abuse [9]. Recently, OSM introduced a daily rate limit in response to a particularly severe vandalism case, marking a shift towards proactive vulnerability mitigation [10]. This paper documents and analyzes the events that led to the implementation of the rate limit, aiming to understand better through these events the factors influencing OSM’s resilience and vulnerability. The chain of events began on October 20, 2023, when members of the OSM_Israel Telegram group [11] reported an instance of vandalism. By the following evening, it became apparent that this was part of a coordinated effort involving newly created accounts Grinberger, A. Y., & Minghini, M. (2025). What happens when VGI is threatened? An analysis of the events behind the introduction of rate limiting in OpenStreetMap. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.). Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024. Available at https://zenodo.org/communities/osmscience-2024. DOI: 10.5281/zenodo.17369467 © 2025 by the authors. Available under the terms of the Creative Commons Attribution (CC BY 4.0) license. 37 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya systematically deleting and distorting data, e.g. [12,13]. Despite DWG’s prompt response with bans, new accounts continued to emerge. The timing of these edits, coinciding with Israeli military’s airstrikes on Gaza in response to Hamas’s October 7 attack, along with changeset comments such as “There is no country Israel & Free Palestine,” clearly indicated a political motivation behind the vandalism. Eventually, 13 new accounts (identified here based on bans and changeset comments) responsible for 279 changesets had succeeded in deleting large portions of Tel Aviv (see Fig. 1a), distorting entities across Israel (see Fig. 1b), and even altering coastlines in Lebanon, Turkey, Cyprus, Greece, Egypt, and the Gaza Strip itself, likely due to the automated edits not excluding entities extending beyond Israel, such as the Levantine Sea [14] (see Fig. 1c). One unique case of counter-vandalism was registered when part of the Gaza strip was annexed into Israel “[i]n response to the deletion of Tel Aviv” [15]. Figure 1. Screenshots of the OSM website, 25 October 2024. (a) Data deletion in Tel Aviv’s center; (b) Northern Tel Aviv - diagonal streets are the result of distortions, the orange line is Israel’s border which was also distorted; (c) Beirut, Lebanon, near the US embassy - the coastline (in orange) crosses to land at one point, the international border is also distorted. Source: ©OpenStreetMap contributors. A secondary chain of events unfolded in response to these attacks, characterized by collaboration between the Israeli mapping community and the DWG. Despite historical tensions between the Israeli community and the DWG [3], a close partnership emerged. On October 22, 2024, a DWG member initiated a dedicated thread on the Israel OSM Community Forum [16] and was subsequently invited to join the OSM_Israel Telegram group, establishing two direct communication channels for reporting new vandal accounts and coordinating data reverts. A clear division of labor emerged: users reported issues, the DWG issued bans and implemented reverts (one highly active Israeli user was unique in also engaging with the reverting processes). The collaboration however extended beyond immediate damage control. Responding to local mappers’ comments in both channels, the DWG member referenced a GitHub issue regarding the possibility of limiting edit volumes [17], explicitly encouraging community members to contribute their thoughts. Although the idea of a rate limiting was not new — the issue had been open since August 2019 — it gained traction after two Israeli community members joined the discussion on October 23, 2023, posting comments which directly addressed the situation in Israel, with one referencing the 38 Proceedings of OSM Science 2024 September 6-8, 2024 | Nairobi, Kenya Assessing the attribute accuracy and logical consistency of road data in OpenStreetMap Wangshu Wang1,2,* and Alexander Zipf2 1 Chair of Cartography and Visual Analytics, Technical University of Munich, Munich, Germany; [email protected] 2 GIScience Research Group, Heidelberg University, Heidelberg, Germany; [email protected] * Author to whom correspondence should be addressed. This abstract was accepted to the OSMScience 2024 Conference after peer-review. OpenStreetMap (OSM) relies on crowdsourced contributions and lacks strict quality control, ensuring data quality has thus become a key area of research [1]. Understanding and addressing these data quality issues can facilitate unlocking OSM's full potential for diverse applications. OSM data quality assessment methods can be divided into two broad categories: extrinsic and intrinsic. Extrinsic quality assessment methods compare OSM data with a reference dataset (e.g., authoritative data sources). This is where most initial research on OSM data quality started from [2-4]. Yet, a reference dataset may not always be available. On this ground, researchers called for attention to intrinsic indicators of OSM data quality [5] and proposed intrinsic data quality measures based on a snapshot of data, data history, and metadata [5-9]. It is crucial to acknowledge that the quality of OSM data is not a one-size-fits-all metric. Rather, it heavily depends on the purpose of the application domain, known as "fitness-for-use" [5,10]. The diverse application potentials introduce dynamic data requirements. Navigation is one of the primary application domains in which OSM plays a pivotal role. Among them, vehicular traffic constitutes a significant portion of road usage and has a substantial impact on urban mobility and infrastructure planning. Additionally, the complexities involved in automotive navigation and traffic systems require a higher level of data accuracy and reliability, making it a compelling starting point for investigating intrinsic road data quality. In the context of car driving, the attribute accuracy and logical consistency of road data are particularly important [11]. Attribute accuracy refers to the correctness and logical coherence of attributes associated with road features, such as speed limits, road classifications, and turn restrictions [12]. Logical consistency ensures that the road network data follows the correct topological rules, such as proper connectivity of different road classes. The attribute accuracy and logical consistency of OSM road data are essential for traffic planning and supporting navigation applications. To address the challenges of evaluating OSM data quality for navigation, when a reference dataset is unavailable, this research narrows its focus to the attribute accuracy and logical consistency of OSM road data. Our current work in progress uses the Munich city centre as a case study. Wang, W., & Zipf, A. (2025). Assessing the attribute accuracy and logical consistency of road data in OpenStreetMap. In: Minghini, M., Grinberger, A. Y., & Mooney, P. (Eds.). Proceedings of OSM Science 2024, Nairobi, Kenya, 6-8 September 2024. Available at https://zenodo.org/communities/osmscience-2024. DOI: 10.5281/zenodo.17369559 © 2025 by the authors. Available under the terms of the Creative Commons Attribution (CC BY 4.0) license. 45 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya To assess attribute accuracy, we first identify important attributes related to car driving, such as speed limits, number of lanes, and one-way information [13-15]. Then, our method assesses the attribute completeness of each segment as the weighted sum of the value of each attribute divided by the weighted sum of each attribute with its maximum value. The attribute completeness provides us with a first overview of the attribute data. For instance, as of June 30, 2024, the completeness level of the Munich city centre is 67.185%. In the next step, a set of rules based on country-specific traffic laws is defined, and the relevant attributes are checked against these rules. So far, we have looked into the traffic law in Germany [16] and the aggregated default speed limits on the OSM wiki [17]. We defined rules accordingly for speed limits regarding the “highway” type. To determine whether a road was in an urban or rural area, relevant attributes like “zone:maxspeed” “source:maxspeed”, “maxspeed:type”, “zone:traffic” were checked. Then the algorithm verified the value of road segments against the rules. When a mismatch occurred, these road segments were identified and considered as inaccurate. The inaccuracy of the speed limit is calculated as the length of road segments with an inaccurate “maxspeed” value divided by the total road length with the “maxspeed” attribute. Our case study in the Munich city centre revealed a low attribute inaccuracy rate of 0.078%. However, some discrepancies were detected, such as residential roads with a speed limit of 60 km/h (see Figure 1). Figure 1. An example of a detected road segment with inaccurate maxspeed value. It is marked as a residential road while having a maxspeed of 60 km/h. Regarding logical consistency, a set of rules was defined based on country-specific conventions, and inconsistent cases were detected against these rules. We acknowledge that different regions in the world have different road construction and mapping conventions. So far, we have conducted our case study in Germany, and defined rules for the values of the “highway” tag. These rules include: the connection of different classes of roads should be logically consistent (i.e., a way with a high level of importance in the road network should not be connected directly to a way with a much lower level of importance); 46 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya motorways can only be connected to “motorway” or “motorway_link”; a link road should be connected to its corresponding highway. For each inconsistency type, we calculated the inconsistency rate as the length of affected roads divided by the total length. Since we considered the aforementioned inconsistency the same weight, the logical inconsistency was then calculated as the sum. As such, we are flexible in adding new rules to the indicator. For example, regarding the aforementioned type of logical inconsistency, the Munich city centre has rates of 0.916%, 0.106%, and 0.004%, respectively. The overall logical inconsistency rate is 1.026%. Our work aims to assess OSM road data quality with a focus on car driving. Current work-in-progress proposed indicators to assess attribute accuracy, with a focus on speed limits and methods for estimating the logical consistency of road data. Checking the speed limits and their sources against default speed limits offers an initial assessment to attribute inaccuracy, as the speed limit is one of the most important road attributes related to car driving. It also provides the starting point of a scalable approach to extend to a global scale. Worth noting for attribute inaccuracy is that this indicator assesses the data against the country's default speed limits. In reality, many roads are regulated by traffic signs, with more strict speed limits. If a road segment is not flagged as inaccurate, it does not necessarily mean it is accurate. This indicator shows the road attribute quality, but should not be understood as indicating accuracy. In addition, exceptional cases may exist, for different temporal scales, which should also be treated cautiously. As an indicator aggregated by individual measures, logical consistency offers varying levels of detail, thus suitable for different levels of decision-making. The logical inconsistency rate presents a high-level overview of the data quality. Users can make informed decisions on whether to use OSM as a data source for specified areas. Once the high-level decision is made, examining the error types and searching for relevant solutions to process data can be the next step. When the inconsistency rate is low and occurs at the individual case level, reviewing the detected cases is more beneficial than focusing on the calculated inconsistency rate. In the next step, we plan to conduct a user study with navigation service providers, aiming at identifying crucial attributes for routing services tailored to car driving, in order to improve the attribute completeness indicator. In terms of attribute inaccuracy, we plan to scale it up to the global level, by implementing the default speed limits aggregated from contributors on the OSM Wiki [17], as well as using large language models to help verify the aggregated speed limits. As exception cases for speed limits exist in reality, we can consider introducing other tags into the analysis. Another possibility would be leveraging the crowd wisdom further to collect exceptional cases, by potentially collecting such data in the OSM wiki page. With the proposed assessment, OSM data users can verify road data quality in terms of attribute accuracy and logical consistency, before their intended use. With the evaluation method, we also aim to inform the potential improvement of OSM road data quality, especially for navigation. 47 Proceedings of the OSMScience 2024 September 6-8, 2024 | Nairobi, Kenya References [1] Yan, Y., Feng, C., Huang, W., Fan, H., Wang, Y., & Zipf, A. (2020). Volunteered geographic information research in the first decade: a narrative review of selected journal articles in GIScience. International Journal of Geographical Information Science, 34(9), 1765–1791. [2] Girres, J., & Touya, G. (2010). Quality assessment of the French OpenStreetMap dataset. Transactions in GIS, 14(4), 435–459. [3] Haklay, M. (2010). How good is volunteered geographical information? A comparative analysis of OpenStreetMap and Ordnance Survey datasets. Environment and Planning B: Planning and Design, 37(4), 682–703. [4] Zielstra, D., & Zipf, A. (2010). A comparative study of proprietary geodata and volunteered geographic information for Germany. Proceedings of the 13th AGILE International Conference on Geographic Information Science, 1–15. [5] Barron, C., Neis, P., & Zipf, A. (2014). A comprehensive framework for intrinsic OpenStreetMap quality analysis. Transactions in GIS, 18(6), 877–895. [6] Fogliaroni, P., D’Antonio, F., & Clementini, E. (2018). Data trustworthiness and user reputation as indicators of VGI quality. Geo-spatial Information Science, 21(3), 213–233. [7] Nejad, R. G., Abbaspour, R. A., & Chehreghan, A. (2022). Spatiotemporal VGI contributor reputation system based on implicit evaluation relations. Geocarto International, 37(26), 12014–12041. [8] Severinsen, J., De Róiste, M., Reitsma, F., & Hartato, E. (2019). VGTrust: measuring trust for volunteered geographic information. International Journal of Geographical Information Science, 33(8), 1683–1701. [9] Sundaram, R. C., Naghizade, E., Borovica-Gajić, R., & Tomko, M. (2021). Can you fixme? An intrinsic classification of contributor-identified spatial data issues using topic models. International Journal of Geographical Information Science, 36(1), 1–30. [10] Senaratne, H., Mobasheri, A., Ali, A. L., Capineri, C., & Haklay, M. (2017). A review of volunteered geographic information quality assessment methods. International Journal of Geographical Information Science, 31(1), 139–167. [11] Wu, H., Lin, A., Clarke, K. C., Shi, W., Cardenas-Tristan, A., & Tu, Z. (2021). A comprehensive quality assessment framework for linear features from Volunteered Geographic Information. International Journal of Geographical Information Science, 35(9), 1826–1847. [12] Alghanim, A., Jilani, M., Bertolotto, M., & McArdle, G. (2021). Leveraging road characteristics and contributor behaviour for assessing road type quality in OSM. ISPRS International Journal of Geo-information, 10(7), 436. [13] Hiller, J., Müller, F., & Eckstein, L. (2021). Aggregation of Road Characteristics from Online Maps and Evaluation of Datasets. Proceedings of 2021 IEEE Intelligent Vehicles Symposium (IV), Nagoya, Japan, 208–214. [14] Liu, Y., & Hansen, J. H. (2019). Towards complexity level classification of driving scenarios using environmental information. Proceedings of 2019 IEEE Intelligent Transportation Systems Conference (ITSC), Auckland, New Zealand, 810–815. [15] Zheng, Y., Izzat, I. H., & Hansen, J. H. (2019). Exploring OpenStreetMap availability for driving environment understanding. arXiv preprint, arXiv:1903.04084. [16] Straßenverkehrs-Ordnung (StVO) (2024). Retrieved from https://www.gesetze-im-internet.de/stvo_ 2013/BJNR036710013.html [17] OpenStreetMap Wiki (2024). Default speed limits. Retrieved from https://wiki.openstreetmap.org/ wiki/Default_speed_limits 48