ambiente

M. Minghini* Collaborative mapping L. Delucchi** A. Sarretta*** response to disasters through F. Lupia**** M. Napolitano***** OpenStreetMap: the case of A. Palmas****** the 2016 Italian earthquake * Politecnico di Milano, Department of Civil and Environmental Engineering, Milano, Italy ** Fondazione Edmund Mach, Research Digital humanitarians represent the current generation of volunteers, providing timely contributions and Innovation Centre, Department of in the form of digital map data in the aftermath of natural disasters. Starting from the tragic 2010 Biodiversity and Molecular Ecology, earthquake in Haiti and thanks to the success of the OpenStreetMap (OSM) project, the presence San Michele all’Adige, Italy and coordination of these volunteers have grown incredibly over the past years. This work investiga- *** Consiglio Nazionale delle Ricerche tes the dynamics of the mapping process and the nature of the OSM volunteers who contributed (CNR) – Istituto di Scienze Marine map data after the 2016 earthquake in Central Italy. The analyses show that existing OSM users (ISMAR), Venezia, Italy were the majority of those contributing to the mapping activity, with less edits performed by new **** CREA Research Centre for Agri- users. The collaborative mapping process was efficiently coordinated through a dedicated platform cultural Policies and Bioeconomy, and the area hit by the earthquake was significantly edited in OSM after the disaster. Roma, Italy Keywords: collaborative mapping, disaster management, earthquake, OpenStreetMap ***** Fondazione Bruno Kessler (FBK), Trento, Italy Risposta collaborativa alla mappatura di eventi disastrosi tramite OpenStreetMap: ****** Wikimedia Italia, Milano, Italy il caso del terremoto in Italia del 2016. I digital humanitarians rappresentano la nuova ge- nerazione di volontari, capaci di fornire contributi tempestivi sotto forma di dati geografici all’indo- mani di disastri naturali. A partire dal tragico terremoto di Haiti nel 2010 e grazie al successo del progetto OpenStreetMap (OSM), la presenza e il coordinamento di questi volontari sono cresciuti a myriad of actors and applications incredibilmente negli ultimi anni. Il presente lavoro analizza le dinamiche del processo di mappatura (Mooney and Minghini, 2017). The e la natura dei volontari di OSM che hanno generato dati geografici dopo il terremoto nel Centro open nature of OSM makes it hi- Italia del 2016. Le analisi mostrano che gli utenti OSM già esistenti prima dell’evento hanno contri- ghly suitable for disaster mapping, buito in modo predominante, mentre quelli registrati al progetto dopo l’evento hanno contribuito in as shown by the response after the misura minore. Il processo di mappatura collaborativa è stato efficacemente coordinato attraverso earthquakes in Haiti in 2010 and una piattaforma dedicata e l’area colpita dal disastro è stata editata significativamente in OSM. Nepal in 2015 (see e.g. Soden and Parole chiave: mappatura collaborativa, gestione dei disastri, terremoto, OpenStreetMap Palen, 2014; Poiani et al., 2016). However, OSM mapping in countri- es prone to many natural hazards is 1. Introduction ospatial datasets, which have been challenged, among other reasons, described with a variety of terms by limited broadband connection, Natural disasters have always including crowdsourcing, Voluntee- lack of GPS devices, technical skills marked the history of humankind. red Geographic Information (VGI), and support from organizations, go- Recent years have witnessed an in- Participatory Sensing and Citizen vernments and academia (Latif et crease in both their frequency and Science (See et al., 2016). In the spe- al., 2011). In the context of natural magnitude at a global level, often cific context of disaster management, disasters, the coordination of volun- with tragic consequences when these practices have seen a new pla- teers’ mapping efforts is operated by events strike in developing or under- yer entering the scene: the so-cal- the Humanitarian OpenStreetMap developed countries where autho- led digital humanitarians, i.e. the Team (HOT, https://www.hotosm. ritative geographic information is networks of technology volunteers org), which was formed right after either not available or not accurate crowdsourcing timely information the Haiti earthquake. The HOT’s and up-to-date (Poiani et al., 2016). during crises response (Meier, 2015). However, in contrast to the past, the The main platform around which For understanding the figures in a better activities falling under the umbrella humanitarian efforts revolve before way, please refer to the online version of of disaster management (Hodgkin- and after disasters is OpenStreetMap the article, available on the GEAM jour- son and Stewart, 1991) can cur- (OSM, http://www.openstreetmap. nal website. rently benefit from the flourishing org), the world’s openly-licensed Per una migliore comprensione delle figu- of collaborative mapping practices geospatial database created by vo- re, si rimanda alla versione online dispo- based on citizens’ provision of ge- lunteers, which is nowadays used by nibile sul sito internet della rivista GEAM.

Geoingegneria Ambientale e Mineraria, Anno LIV, n. 2, agosto 2017, 21-26 21 environment

Tasking Manager (TM, http://tasks. locally owned and maintained to missing, to facilitate medical care hotosm.org) is the software platform keep pace with the changing socio- (Feinman, 2014). Similarly, mapa- used everyday by hundreds of volun- physical environment and as a tool thons were organized by the Ame- teers to perform remote mapping by for facing future threats (Shemak, rican Red Cross and HOT to fight digitizing geospatial objects (mainly 2014; Soden and Palen, 2014). the 2014 Ebola outbreak in Western roads and buildings) on top of satelli- Similarly to the work presented in Africa (Moeller and Furhmann, te imagery. This remote work clearly this paper, by extracting information 2015). Recently, a partnership complements the local knowledge from the OSM database, Poiani et al. between HOT and Digital Globe has of people who are physically present (2016) studied how the collaborati- launched a malaria elimination cam- in the disaster area. With these pre- ve mapping was coordinated before paign in Southern Africa, Southeast mises, the purpose of this study is to and after the Nepal earthquake in Asia and Central America (https:// investigate the OSM collaborative 2015. Due to the major dimension www.hotosm.org/projects/malaria_ mapping process that took place after of the event and the coordination elimination_campaign). the earthquake occurred in Central led by HOT, the study highlighted Italy in August 2016. This is done by huge amounts of edits performed by analysing the number, nature, prove- both new OSM mappers and parti- nance and amount of contributions cipants to the so-called mapathons. 3. The Italian case study of the active OSM volunteers. These are short social events whe- The remainder of the paper is re both experienced OSM mappers On August 24, 2016, at 03:36 AM structured as follows. Section 2 exa- and beginners meet and contribute (local time) a ML 6.0 earthquake (MW mines related work in the field of data to the project (Ebrahim et al., 6.0) struck an extensive portion of OSM collaborative disaster mapping. 2016). In a study carried out on 26 the Central Apennines between the Section 3 introduces the Italian case HOT campaigns, Dittus et al. (2016) towns of Norcia and Amatrice (see study and describes the experimental highlighted that most of the contri- Figure 1). The epicentre was located work performed through the analysis butors are already active at the time near the town of Accumoli. The area of the OSM database history and the of the disaster, a small percentage of was struck by several earthquakes in related results. Section 4 concludes experienced mappers are reactiva- historical times (1627, 1639, 1672, the paper and provides recommenda- ted because of the event and first- 1703) (INGV Working Group on tions on how contributors’ response time mappers provide major contri- Amatrice earthquake, 2016). The can be facilitated and increased. butions which are often biased by 2016 earthquake caused 299 victims quality issues. As a further example, and almost 400 injured people. Seve- Palen et al. (2015) studied the OSM ral other shocks occurred during the volunteers’ contributions after the following weeks and other earthqua- 2. Related work onset of the 2013 Typhoon Yolan- kes struck adjacent areas in the Cen- da (or Haiyan) in the Philippines. tral Italy during the following months Started after the Gaza crises in The authors found a 3-fold increase on October 26 (ML 5.4), October 30 2009, the use of OSM for collabo- in the number of mappers since the (ML 6.5) and January 18, 2017 (4 rative disaster mapping has raised Haiti event, which could be explai- main subsequent shocks, ML 5.1, 5.5, interest in the academic community ned by both the growth of the OSM 5.4, 5.0). since the Haiti earthquake in 2010. community and HOT’s organizing The data analysed and discussed In the aftermath of that disaster, the effort, e.g. thanks to the introduc- in this article refer only to the first contribution of volunteers’ mapping tion of the TM that reduced the event (August 24, 2016). In fact, was shown to be pivotal to deliver overlapping and conflict of edits. this was the main driver for the ac- relief efforts in a place where data- HOT’s cooperation with other or- tivation of the Italian OSM com- bases on population, infrastructure, ganizations has led to projects aimed munity’s mapping efforts. The first location and assets were lacking. at preventive mapping in countries edit in the map was made at 05:44 Thanks to the distribution of up-to- that are high vulnerable to epide- AM, while the first message on the date imagery and the mobilization mics, political crises and natural Italian OSM mailing list was sent of remote mappers from around the hazards and where maps do not exi- at 06:56 AM (https://lists.openstre- world, a geospatial infrastructure for st. As an example, Médecins Sans etmap.org/pipermail/talk-it/2016- Haiti was built to ground disaster Frontières, the British and American August/054643.html). The collabo- response (Zook et al., 2010). The Red Cross and HOT have launched rative mapping response started with mapping efforts for disaster relief a project to add 200M addresses to the activation, at 11:02 AM, of the were extended beyond the event by OSM in crowded towns, villages and first task (Project 13) in the OSM HOT, with the aim to make the map refugee camps where digital maps are TM instance managed by Wikime-

22 Agosto 2017 ambiente

Fig. 1. Seismicity registered from August 24 to August 27 in the study area. The two darker red points identify the epicentres of the main shocks at 03:36 AM and 04:33 AM on August 24, 2016. Sismicità rilevata dal 24 al 27 agosto nell’area di studio. I due punti più scuri corrispondono agli epicentri delle scosse più forti delle 03:36 e 04:33 del 24 agosto 2016. dia Italia (the Italian association interaction was activated with the org/planet/full-history), which promoting Wikipedia and the other European Commission COPERNI- contains the full editing history of Wikimedia projects in Italy and, CUS Emergency Mapping Service the OSM database. Thanks to the since 2016, the Italian OSM chap- (http://emergency.copernicus.eu/ Osmium Tool (http://osmcode.org/ ter) to improve the basemap of the mapping/copernicus-emergency-ma- osmium-tool), the OSM planet area with the pre-earthquake infor- nagement-service) with the purpose file was cut both spatially (on the mation (http://osmit-tm.wmflabs. of adding to OSM the information same geographic area of Project 13 org/project/13). This was followed derived from the remote sensing as- in the OSM TM, which also in- by the activation of other two TM sessment of damages on buildings cluded the areas of Project 14 and tasks: Project 14 (http://osmit-tm. and roads published in a specific Project 15) and temporally (on the wmflabs.org/project/14) and Project activity created by COPERNICUS whole year 2016). In other words, 15 (http://osmit-tm.wmflabs.org/ Emergency Mapping Service for the the resulting file contained the full project/15). It must be said that, due earthquake in Central Italy (http:// history from the beginning to the to the small dimension of the area, emergency.copernicus.eu/mapping/ end of 2016 of all the OSM objects the presence of an already well-coor- list-of-components/EMSR177). included in the area of interest, dinated national community and the i.e. all their versions and changes activation of tasks in the Italian TM (in geometry and attributes), the managed by Wikimedia Italia, no ad- 3.1. Methodology user(s) who made the changes and ditional tasks were activated on the the timestamp of each change. This HOT’s TM. Thus, the primary com- The analyses are primarily ba- file was then imported into a Po- munity of OSM volunteers addressed sed on the OSM full history planet stgreSQL database (https://www. was the Italian one. Finally, a strong file (http://planet.openstreetmap. postgresql.org) enabled with the

Agosto 2017 23 environment

PostGIS spatial extension (http:// event who made the first edit inside first earthquake the contributions of postgis.net), using the osm2pgsql the study area (AI). Results show users who were already active before tool (http://wiki.openstreetmap. that the existing users contributing the event (EBA) are less than the org/wiki/Osm2pgsql). The analysis OSM edits in the study area in 2016 contributions of users “attracted” af- was also performed on the OSM tile were about the 90% of the total: they ter the event (EA), the proportion log (http://planet.openstreetmap. included EB (40.1%), EBA (11.7%) is inverted (contributions of EBA org/tile_logs), a text file containing and EA (38.5%). Among the users greater than those of EA) for the the number of visits to each map tile registered to OSM after the event, second earthquake. A possible rea- (https://en.wikipedia.org/wiki/Ti- the majority (36 users, correspon- son might be that EA are mostly not led_web_map). To run the analysis, ding to 7.1% of the total) were AI local people driven by humanitarian an open source Python script na- and only 13 (2.6% of the total) were concerns who after the first event med OsmEventAnalyst was crea- AO. This shows an overall positive were more motivated than EBA in ted (https://github.com/osmItalia/ response to the event from the OSM contributing to the mapping activi- OsmEventAnalyst). In parallel, user Italian community. ty. Conversely, the second event has contributions from the TM of Wiki- Figure 2 (a) shows the evolu- attracted a higher number of local media Italia were also investigated. tion of the daily number of OSM contributors. users performing edits in the study The results highlight that the area after the earthquake of August collaborative mapping effort was 3.2. Results and discussion 24, 2016. A peak is clearly visible driven by already experienced on August 24 and 25, followed by users. Conversely, new users had Throughout 2016, 506 OSM about two weeks of intense OSM probably not enough experience users have made at least one edit mapping. After that, the activity and practice to be able to signifi- in the study area. These users were remained almost silent until the cantly contribute to the mapping divided into five classes according second earthquake (October 30) activities. To analyse the mapping to their OSM history: existing users when another peak is visible. Fi- performed through the TM of Wi- who modified the area only before gure 2 (b) shows instead the daily kimedia Italia, the lists of OSM the event (EB), existing users who number of OSM edits according to contributors available for Projects modified the area both before and the user categories presented abo- 13, 14 and 15 were first extracted. after the event (EBA), existing users ve. Most of the edits were clearly These include only the contribu- who modified the area only after the performed by the OSM users alrea- tors who have marked as complete event (EA), users registered to OSM dy registered before the earthquake at least one of the sub-areas of each after the event who made the first (EBA and EA), while little con- Project (see http://learnosm.org/ edit outside the study area (AO), tributions were provided by new en/coordination/tasking-manager and users registered to OSM after the users. Interestingly, while for the for details), i.e. they do not include

Fig. 2. Daily number of users editing OSM in the study area (a); stacked daily number of OSM edits per categories of users after the earthquake of August 24, 2016 (b). Numero giornaliero di utenti che hanno editato OSM nell’area di studio (a); numero giornaliero di modifiche a OSM per categorie di utenti dopo il terremoto del 24 agosto 2016 (b).

24 Agosto 2017 ambiente all the contributors who have made 4. Conclusions References at least one edit in the study area. The latter are traditionally much Over the recent years, OSM has Dittus, M., Quattrone, G. and Capra, more than those who have marked become a catalyst for the collabora- L., 2016. Analysing volunteer enga- as complete the project sub-areas. tive mapping response after natural gement in humanitarian mapping: Comparing the OSM usernames of disasters. This work has investigated building contributor communities at the 94 contributors who marked as the response of OSM volunteers af- large scale, in: 19th ACM Conferen- complete the sub-areas of Projects ter the earthquake events occurred ce on Computer-Supported Co- 13, 14 and 15 with the usernames of in Italy in 2016. The analyses show operative Work & Social Compu- the contributors extracted from the that these events have mainly sti- ting, February 27-March 2, 2016, OSM full history planet file, results mulated the existing OSM users to San Francisco (US), ACM, pp. 108- show that the 94 TM users included improve the quantity and quality 118. most of the users active in the af- of data in the event area, while few Ebrahim, M., Minghini, M., Molinari, termath of the first earthquake (to new users have contributed to the M.E. and Torrebruno, A., 2016. Mi- which Projects 13, 14 and 15 were mapping efforts. In addition, the niMapathon: Mapping the world at related) and overall they contribu- study has shown that the area hit 10 years old, in: 8th Annual Inter- ted the 63% of the total OSM map by the earthquakes has gained a lot national Conference on Educa- edits in the study area during the of OSM-related interest in terms of tion and New Learning Technolo- whole 2016. Taking into account both visualization and new data ad- gies (EDULEARN 2016), July 4-6, that the contributors active throu- ded after the disasters. 2016, Barcelona (Spain), IATED, gh the TM were more than the 94, A number of lessons were learned pp. 4200-4208. this confirms that the TM is a very from this experience. First, the availa- Feinmann, J., 2014. How MSF is map- important driver and aggregator for bility and activation of a TM instan- ping the world’s medical emergency disaster-related mapping. ce (like the one of Wikimedia Italia zones. BMJ. 349:g7644. The analysis of the OSM data in in the case study analysed) is crucial Hodgkinson, P.E. and Stewart, M., the study area reveals that most of to coordinate volunteers’ efforts and 1991. Coping with catastrophe: A the objects (~90%) were not mo- especially to prevent the conflicts handbook of disaster management, dified after the earthquakes and the traditionally generated when many Routledge, London. objects with more changes are linear users edit the same area in the after- INGV Working Group on Amatrice data, followed by point and polygon math of a disaster. The availability of earthquake, 2016. First summa- data. Almost all the edited objects post-event satellite imagery is crucial ry report on the 24 August, 2016, were created before the first event, as well to increase the accuracy, the Amatrice earthquake in Central which is in line with the fact that up-to-dateness and, in general, the Italy. doi:10.5281/zenodo.61121 they were edited afterwards to re- fitness-for-use of the OSM data cre- Latif, S., Islam, K.M.R., Khan, M.M.I. and flect the damages or changes caused ated/edited by volunteers. After the Ahmed, S.I., 2011. OpenStreetMap by the earthquakes. The number of Italian earthquake, OSM data were for the disaster management in OSM objects with changes in geo- used on the field by rescue workers, Bangladesh, in: 2011 IEEE Confe- metry was approximately the double fire fighters and the Civil Protection. rence on Open Systems, Septem- of those with changes in the attribu- Despite this is a very successful re- ber 25-28, 2011, Langkawi (Malay- tes (or tags). Finally, the analysis of sult, in a future outlook the situation sia), IEEE, pp. 429-433. the OSM tile log files in the study might be still improved, e.g. by fur- Meier, P., 2015. Digital Humanitarians. area shows an expected peak on Au- ther strengthening and coordinating How Big Data Is Changing the Face gust 24, 2016. Visualization of the the OSM national community; by of Humanitarian. Boca Raton, Lon- tiles corresponding to the city centre encouraging institutions to make use don and CRC Press, New York. of Amatrice (WGS84 lon-lat coor- of the OSM basemap on their official Moeller, M.S. and Furhmann, S., 2015. dinates 13.2891755, 42.6292367) websites to advertise and promote the Mapping the world: A new approach reveals that before the event the project, to adopt policies based on for Volunteered Geographic Informa- area was almost not visited at all in releasing open geospatial data (not tion in the cloud. The International OSM, while it was still visited after only after disasters) and enabling Archives of Photogrammetry, Re- months from the event. their reuse in OSM; and by enhan- mote Sensing and Spatial Informa- The detailed results of this cing the contacts between the OSM tion Sciences. XL-6/W1. pp. 9-10. analysis on OSM data, including ad- community, the Civil Protection and Mooney, P. and Minghini, M., 2017. A ditional plots and graphs, are availa- the Copernicus Emergency Manage- Review of OpenStreetMap Data, in: ble at http://www.geodati.fmach.it/ ment Service to optimize the chan- Foody, G., See, L., Fritz, S., Mooney, osm_paper_geam.html. nels for data/information exchange. P., Olteanu-Raimond, A.-M., Fonte,

Agosto 2017 25 environment

C.C. and Antoniou, V., Eds), Map- Web references EMSR177 (accessed 07/07/2017) ping and the Citizen Sensor, Ubi- http://planet.openstreetmap.org/ quity Press, London, pp. 37-59. http://www.openstreetmap.org planet/full-history (accessed Palen, L., Soden, R., Anderson, T.J. and (accessed 07/07/2017) 07/07/2017) Barrenechea, M., 2015. Success & https://www.hotosm.org (accessed http://osmcode.org/osmium-tool scale in a data-producing organiza- 07/07/2017) (accessed 07/07/2017) tion: The socio-technical evolution of http://tasks.hotosm.org (accessed https://www.postgresql.org (acces- OpenStreetMap in response to hu- 07/07/2017) sed 07/07/2017) manitarian events, in: 33rd annual https://www.hotosm.org/projects/ http://postgis.net (accessed ACM conference on human fac- malaria_elimination_campaign 07/07/2017) tors in computing systems, April (accessed 07/07/2017) http://wiki.openstreetmap.org/ 18-23, 2015, Seoul (Republic of http://emergency.copernicus.eu/ wiki/Osm2pgsql (accessed Korea), ACM, pp. 4113-4122. mapping/copernicus-emergency- 07/07/2017) Poiani, T.H., dos Santos Rocha, R., De- management-service (accessed https://en.wikipedia.org/wiki/Tiled_ grossi, L.C. and de Albuquerque, 07/07/2017) web_map (accessed 07/07/2017) J.P., 2016. Potential of collaborative https://lists.openstreetmap.org/piper- http://planet.openstreetmap.org/ mapping for disaster relief: A case mail/talk-it/2016-August/054643. tile_logs (accessed 07/07/2017) study of openstreetmap in the ne- html (accessed 07/07/2017) https://github.com/osmItalia/ pal earthquake 2015, in: 49th Ha- http://osmit-tm.wmflabs.org/ OsmEventAnalyst (accessed waii International Conference on project/13 (accessed 07/07/2017) 07/07/2017) System Sciences (HICSS), Janua- http://osmit-tm.wmflabs.org/ http://learnosm.org/en/coordina- ry 5-8, 2016, Koloa, HI (USA), project/14 (accessed 07/07/2017) tion/tasking-manager (accessed pp.188-197. http://osmit-tm.wmflabs.org/ 07/07/2017) See, L., Mooney, P., Foody, G., Bastin, L., project/15 (accessed 07/07/2017) http://www.geodati.fmach.it/ Comber, A., Estima, J., Fritz, S., Ker- http://emergency.copernicus.eu/ osm_paper_geam.html (accessed le, N., Jiang, B., Laakso, M., Liu, H.- mapping/list-of-components/ 07/07/2017) Y., Milčinski, G., Nikšič, M., Painho, M., Pődör, A., Olteanu-Raimond, A.-M. and Rutzinger, M., 2016. Crowdsourcing, citizen science or volunteered geographic information? The current state of crowdsourced geographic information. ISPRS In- ternational Journal of Geo-Infor- mation. 5(5), 55. Shemak, A., 2014. The cartographic dimensions of humanitarianism: Mapping refugee spaces in post- earthquake Haiti. Cultural Dyna- mics. 26(3). pp. 251-275. Soden, R. and Palen, L., 2014. From crowdsourced mapping to commu- nity mapping: The post-earthquake work of OpenStreetMap Haiti, in: 11th International Conference on the Design of Cooperative Systems, May 27-30, 2014, Nice (France), pp. 311-326. Zook, M., Graham, M., Shelton, T. and Gorman, S., 2010. Volunte- ered geographic information and crowdsourcing disaster relief: a case Acknowledgements study of the Haitian earthquake. World Medical & Health Policy. The authors would like to thank the Italian OSM community and particularly 2(2). pp. 7-33. those contributors active after the earthquakes.

26 Agosto 2017