Remote Sensing Based Post-Disaster Damage Mapping – Ready for a Collaborative Approach?

EarthzineArticles, Disaster Management Theme, Earth Observation, Original

Example damage maps of parts of Port-au-Prince (Haiti) following the 2010 earthquake.

Norman Kerle
Faculty of Geo-information Science and Earth Observation (ITC) of the University of Twente, Dept. of Earth Systems Analysis,
Hengelosestraat 99, P.O. Box 217, 7500 AE Enschede, The Netherlands
Rapid and accurate assessment of structural damage is essential after disaster events, especially in densely built-up urban areas. The results provide guidance for rescue forces and other immediate relief efforts, as well as subsequent rehabilitation and reconstruction. Especially for spatially extensive events, ground-based mapping is too slow, typically hindered by disaster-related site access difficulties, or too dangerous as in the case of potential radioactive contamination resulting from the March 2011 earthquake and tsunami disaster in Japan. Remote sensing has long been seen as a potential solution. Essentially, any type of platform-sensor combination is capable of providing some form of useful view on a disaster scene. However, optimal remote sensing strategies that consider the actual information needs of specific stakeholders, the technical utility of a given platform and sensor with respect to the types of infrastructure and damage present, as well as temporal and environmental limitations, are harder to determine. The utility of airborne and spaceborne remote sensing in emergency response was reviewed by Kerle et al. [1] and Zhang and Kerle [2], respectively.
Many studies have investigated the utility of imagery for damage mapping, covering the entire spectrum from low-cost and uncalibrated airborne data (still or video imagery), deployed on balloons, kites, unmanned aerial vehicles or piloted aircraft, to sophisticated multi- or hyperspectral, lidar, thermal or radar devices mounted on air- or spaceborne platforms. To respond rapidly to a disaster anywhere in the world, often only satellites offer a solution. To realize this potential, and to create a dependable disaster support instrument, the International Charter “Space and Major Disasters” was set up by the European, French and Canadian space agencies [see 3], and has been activated nearly 300 times since November 2000. The Charter is only responsible for the initial satellite image acquisition, while the subsequent data processing and damage mapping is mainly done by the DLR Center for Crisis Information (DLR-ZKI), UNOSAT, or the Service Régional de Traitement d’Image et de Télédétection (SERTIT, based at Strasbourg University, France).
While damage map generation based on Charter data has long been a routine activity, recent disaster events have highlighted (i) limitations in the process, (ii) newly emerging and potentially competitive methods, and (iii) a far wider field of stakeholders (both map producers and users) than ever before. At the same time we are seeing a democratization of geodata and –tool access, which, together with growing spatial literacy outside the professional geoinformatics domain, also offers the potential for more collaborative damage mapping. The suitability and potential challenges of such approaches are addressed in this article.
Traditional image-based damage mapping
The Charter is generally seen as a successful example of international cooperation [4], and also the resulting damage mapping has been repeatedly praised [e.g. 5, 6]. Others, however, have been more critical. Inglada and Giros [7] remarked that map quality needs to be given more attention in the overall assessment of Charter success. However, the increasing time pressure of Charter activations that have reached an average of 40 per year, the growing number of maps produced per disaster, the typical lack of ground information, and the largely Europe-based map production, mean that map accuracies are rarely determined. Kerle [8] assessed the accuracy of different damage map types produced after the 2006 Indonesia earthquake by UNOSAT and DLR-ZKI. They depicted damage interpolated per grid cell (DLR-ZKI), and as building- and block-level polygons (UNOSAT), of which only the coarser damage polygons were very accurate. While the variable map accuracy can be attributed to a number of factors, such as image type and cloud contamination, it must be viewed in the context of a wider set of problems:

• Universally accepted damage map nomenclature and style are lacking;
• Damage is depicted on various scales and in different categories, using point or line signatures, damage clusters, grid-based damage averages, damage per city block using color ramps, damage aggregated per neighborhood, or as continuous damage density maps (Figure 1);
• The decisions for a given mapping style do not appear to be based on what users have identified as useful, or to reflect the needs of specific user groups;
• A growing number of organizations produce damage maps, including non-experts, leading to duplication of mapping efforts and potential disagreement;
• The number of damage map users, and their information needs, have been growing rapidly;
• Traditional Charter-maps remain static, being distributed as print-optimized pdf documents, not allowing ready mash-ups with other data and map customization; and
• Damage mapping validation rarely takes place.

This leads to the interesting situation where, on one hand, the Charter process is well-established and increasingly taken for granted. However, it also appears that the map-making process needs to be reevaluated and aligned with actual multi-stakeholder information and map-type needs (both static and dynamic), coupled with a more consistent damage map nomenclature and style. The latter needs to be done based on both scientific analysis of previous damage map results and a thorough understanding of the user community and its diverse needs.

Example damage maps of parts of Port-au-Prince (Haiti) following the 2010 earthquake.

Figure 1 . Example damage maps of parts of Port-au-Prince (Haiti) following the 2010 earthquake; a- SERTIT, b- ITHACA, c- UNSC, d- iMMAP, e- DLR-ZKI, f- e-GEOS; legends adapted from original sources to show only elements related to structural damage. Map scales are variable. Location of Figure 2 is shown in a.

Recent trends in disaster damage mapping
Maps by and for the masses
With the digital age, cartography has moved beyond its traditional boundaries, developing from a highly accurate and tightly defined craft into a core component of a growing number of enabling technologies [9]. This essentially allows anyone to be a map maker (though not necessarily a cartographer), and makes anyone with a digital media device a likely map user, a trend evident in disaster damage mapping. On the “official” damage mapping side, the Charter-data processing agencies used to be the only major damage map makers [10]. The 2010 Haiti earthquake demonstrated how much that has changed, with remote sensing-based maps also being made by Information Technology for Humanitarian Assistance, Cooperation and Action (ITHACA, based at Torino University, Italy), the United Nations Cartographic Section (UNCS), the European Union Satellite Center (EUSC), Information Management & Mine Action Programs (iMMAP), the Joint Research Centers (JRC), and an alliance of the Italian Space Agency and Telespazio that produces damage maps under its own e-GEOS label, as well as that of its GMES service G-MOSAIC (see Figure 1). Also commercial players, such as the ExpressMaps service by SPOT Imaging and Infoterra France, provided reference maps [8]. In principle, such an extensive mapping is good, given the growing number of disaster response agencies (in the Haiti earthquake aftermath, up to 10,000 NGOs were estimated to be active) who require ever more detailed information for more effective and coordinated work. In reality, however, the more than 2000 damage maps for Haiti catalogued by Reliefweb suggest considerable duplication and a lack of coordination. Which of those maps were actually used (and found useful) also remains to be assessed.
Neogeography & collaboration
Interaction and collaboration are key terms in the Web 2.0 philosophy that are increasingly moving into scientific territory. However, the roots of citizen science, i.e. the contribution of the wider population to answer scientific questions, already date back to the year 1900 (an Audubon Society bird counting program). It can comprise contributors who may or may not be lay persons, and who either actively contribute to answering the scientific problem at hand, or merely provide distributed computing resources (such as the Africa@home, or the better-known SETI@home projects). In the disaster arena, it is a more recent phenomenon, but one that is quickly establishing itself in a number of different forms, as recently reviewed by Goodchild and Glennon [11]. People serve as distributed sensors, such as for the USGS’s “Did You Feel It?” to map seismic intensities, or neogeography tools [12] are used to aggregate and visualize community damage reports, such as after the 2011 Queensland (Australia) floods. Ushahidi and tools such as Crowdmap provide similar platforms for such information integration. At its simplest, geotagged photos provide a rough overview of a disaster site. However, the geotools by Google have had the most profound impact on damage mapping. Google Earth has clearly resulted in a hugely enlarged spatial literacy and wider interest in geodata, led to a democratization of geodata access, and served as an effective and collaborative platform for distribution or image data and derivatives. This interest, as well as the growing strength of the citizen mapper, is well reflected in the Crisismappers community that now has more than 1,000 members, and that largely took charge of coordinating the unofficial Haiti damage mapping.
Collaborative damage mapping
Following the Haiti earthquake, there were two prominent approaches to post-disaster mapping using Google tools. The most visible was Google Map Maker (GMM; and its open-source equivalent Open Street Map, OSM), whose rapid mapping of Port-au-Prince has been well documented [e.g. 10]. Hundreds of people with local knowledge created a comprehensive basemap of the disaster area within a few days, working on image data but often also using ground knowledge. Following the Wiki principle, anyone was allowed to map new or to edit existing map elements, with a moderation component ensuring accuracy and error correction. This approach may not meet all traditional cartographic benchmarks, but results from Haiti have indicated high coverage and geometric accuracy [10]. The potential of such base data mapping also became clear quickly, with many other services being built around it (see, such as damage mapping or the OpenRouteService to map road accessibility.
The second form of collaborative mapping was done under the Global Earth Observation-Catastrophe Assessment Network (GEO-CAN) initiative led by the World Bank and the Global Facility for Disaster Reduction and Recovery (GFDRR), and carried out solely using remote sensing data. More than 500 individuals mapped damage visually, identifying D4 and D5 buildings (very heavy damage and complete destruction, respectively, on the European Macroseismic Scale of 1998). Based on experiences following the 2008 Wenchuan (China) earthquake, where image-based collaborative mapping had been done using the Virtual Disaster Viewer developed by ImageCat, Google Earth was used as the platform for Haiti.. Unlike the collaborative basedata mapping in GMM and OSM, the mapping was done by remote sensing experts. Initially, 50cm Geoeye imagery was used to derive point locations of collapsed buildings. This was followed by a more detailed and extensive mapping on 15 cm aerial imagery, using polygons to outline D4 and D5 buildings. The work was distributed through grid cells, for which mappers submitted kmz files of damage outlines that were later integrated at ImageCat (Figure 2a). However, how those outlines were specifically processed and later became part of the Joint Remote Sensing Damage Assessment Database of UNOSAT, World Bank and the JRC (Figure 2b), is not clear.
Example of collaborative damage mapping results based on 15 cm aerial imagery by expert volunteers organized by ImageCat for parts of Port-au-Prince.

Figure 2. a- Example of collaborative damage mapping results based on 15 cm aerial imagery by expert volunteers organized by ImageCat for parts of Port-au-Prince (see Figure 1 a for location), and b- the damage information from the Joint Remote Sensing Damage Assessment Database of UNOSAT, World Bank and the JRC for the same area, based on both aerial and satellite data. How the ImageCat data were integrated with other image-based mapping results is unclear, in particular to what extent errors were detected, disagreements resolved, or different damage classifications were reconciled (e.g. yellow boxes).

Is collaborative damage mapping the way forward?
Image-based damage assessment can be considered more challenging than base data mapping (roads, bridges, etc.), explaining why the GEO-CAN mapping only relied on expert contributions. However, even a basic level of consistency can only be achieved if both mapping goal and strategy are clearly explained [13]. For Haiti, detailed instructions were provided that explained the damage to be mapped and gave example illustrations. The accuracy of that mapping effort is not known, with validation work still ongoing by ImageCat, the Joint Research Centers (JRC) and UNOSAT. However, visual damage mapping results for recent disasters have been relatively sobering. The working assumption used to be that with higher spatial resolution structural damage mapping would become increasingly accurate, and to some extent that has been true. With modern very high spatial resolution data, such as from Worldview-2 or Geoeye-1 (0.5 and 0.41m panchromatic resolution, respectively) identification of individual buildings has become possible, and the detection of collapsed buildings more accurate. However, even at such level of detail identification of lower damage scales remains difficult [14]. Also in Haiti, it was found that building damage mapped in the 15 cm aerial imagery was approximately 10 times higher than what had been identified in the 50 cm satellite data [15]. Even visual multi-view damage mapping proved challenging. Cambridge Architectural Research Ltd. carried out visual damage mapping on airborne Pictometry data (imagery that includes 4 oblique views and a vertical image) acquired shortly after the Haiti earthquake. Despite the superb data quality, only about 63% of the buildings mapped as D4 and D5 on the ground for validation purposes were identified as such in the visual analysis of the Pictometry imagery (K. Saito, pers. comm.).
Open access or closed systems?
Web 2.0 technologies allow a number of ways to support collaborative mapping, with suitability dictated by the mapping aims, the qualifications and spatial literacy of the mapper, and confidentiality and data access restrictions. Early tools for post-disaster collaboration tended to be of the sophisticated GIS and web map server technology, including both proprietary and open-source systems. Many are meant to support multi-stakeholder decision making, such as the Distributed Virtual Geographic Environment [DVGE; 16] or the User-Defined Operational Picture [UDOP; 17], built for the U.S. Southern Command to facilitate Google Earth-based data integration and decision making for the Haiti disaster response. Others developed more rigid Geo Web Services to share damage mapping results and allow controlled interaction with registered users, e.g. to resolve mapping ambiguities [18]. Both system types can serve a specific purpose well; however, with system transparency and ease of use being critical components of successful collaborative mapping [13], tools based on familiar technology of the Google type can more easily reach a large number of contributors.
Design of a more robust collaborative mapping strategy
The collaborative image-based mapping employed in the GEO-CAN effort is attractive, since it can facilitate an efficient and coordinated parallel image analysis by hundreds of experts. The controlled access conditions also allow tighter control over the instructions issued to the group and over the integration of submitted results than is possible with GMM or OSM, and every damage element mapped is clearly linked to a specific contributor. However, a number of technicalities need to be resolved before the suitability of the approach can be assessed: (i) how should mapping instructions best be crafted and presented to avoid ambiguities and ensure consistent mapping?, (ii) how can an optimal balance be achieved between higher mapping detail (e.g. annotated polygons vs. simple damage points) and accuracy and consistency declining with increasing mapping complexity?, (iii) should the same tile be mapped by several experts to provide a consensus and highlight mapping errors?, and (iv) how can the accuracy of the results be objectively assessed, and remaining errors and ambiguities identified? Whatever the strategy, it is critical that the mappers are provided with feedback on their work and the results [19].
What do the three distinct damage mapping approaches – (i) static maps based on Charter data (satellite imagery), (ii) collaborative grid-based damage mapping by many experts (satellite and aerial images), and (iii) the mapping largely done by lay person in GMM and OSM (using both ground knowledge and remote sensing images) – teach us? The rapid and flexible mapping response in the aftermath of the Haiti earthquake carried out outside the official image-based damage mapping domain has not only created competition for traditional Charter-based map products. It has also shown that the technology to support collaborative mapping is mature, with the seemingly most robust and accessible platform having emerged from open access virtual globes such as Google Earth, rather than from more traditional web map services previously explored by disaster response researchers. The methods developed or refined for Haiti have shown that (i) considerable potential in post-disaster information gathering, including of ground-based damage evidence, also lies with non-professionals, (ii) there is a great willingness by non-mandated people to contribute to such efforts, and (iii) there is a great need for flexible and customized map products, and that those can be readily provided. The traditional Charter-type mapping process needs to move away from static, one-directional mapping, and strive for a better understanding of map user needs, as well as damage maps with standardized and agreed-upon nomenclature.
However, actual image-based structural damage mapping by lay persons is not advisable. Here an expert-based collaborative approach is more useful, though an optimal methodology both for the mapping and the subsequent data integration and processing has yet to be developed. This process needs to be transparent, and feedback must be given to the mappers. It also needs to be clearly communicated to potential users that map accuracies will be modest only even for high resolution data.
[1] N. Kerle, S. Heuel, and N. Pfeifer, “Real-time data collection and information generation using airborne sensors,” in Geospatial Information Technology for Emergency Response, S. Zlatanova and J. Li, Eds., ed London: Taylor & Francis, 2008, pp. 43-74.
[2] Y. Zhang and N. Kerle, “Satellite remote sensing for near-real time data collection,” in Geospatial Information Technology for Emergency Response, S. Zlatanova and J. Li, Eds., ed London: Taylor & Francis, 2008, pp. 75-102.
[3] J. L. Bessis, J. Bequignon, and A. Mahmood, “Three typical examples of activation of the International Charter “Space and Major Disasters”,” Advances in Space Research, vol. 33, pp. 244-248, 2004.
[4] A. Ito, “Issues in the implementation of the International Charter on Space and Major Disasters,” Space Policy, vol. 21, pp. 141-149, May 2005.
[5] I. Z. Gitas, A. Polychronaki, T. Katagis, and G. Mallinis, “Contribution of remote sensing to disaster management activities: A case study of the large fires in the Peloponnese, Greece,” International Journal of Remote Sensing, vol. 29, pp. 1847-1853, 2008.
[6] S. Voigt, T. Kemper, T. Riedlinger, R. Kiefl, K. Scholte, and H. Mehl, “Satellite image analysis for disaster and crisis-management support,” IEEE Transactions on Geoscience and Remote Sensing, vol. 45, pp. 1520-1528, Jun 2007.
[7] J. Inglada and A. Giros, “On the real capabilities of remote sensing for disaster management – feedback from real cases,” in International Geoscience and Remote Sensing Symposium (IGARSS’04), Anchorage, US, 2004.
[8] N. Kerle, “Satellite-based damage mapping following the 2006 Indonesia earthquake–How accurate was it?,” International Journal of Applied Earth Observation and Geoinformation, vol. 12, pp. 466-476, 2010.
[9] M. Monmonier, “Cartography: uncertainty, interventions, and dynamic display,” Progress in Human Geography, vol. 30, pp. 373-381, Jun 2006.
[10] C. Heipke, “Crowdsourcing geospatial data,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 65, pp. 550-557, Nov 2010.
[11] M. F. Goodchild and J. A. Glennon, “Crowdsourcing geographic information for disaster response: a research frontier,” International Journal of Digital Earth, vol. 3, pp. 231-241, 2010.
[12] S. Elwood, “Geographic Information Science: new geovisualization technologies – emerging questions and linkages with GIScience research,” Progress in Human Geography, vol. 33, pp. 256-263, Apr 2009.
[13] G. Newman, D. Zimmerman, A. Crall, M. Laituri, J. Graham, and L. Stapel, “User-friendly web mapping: lessons from a citizen science website,” International Journal of Geographical Information Science, vol. 24, pp. 1851-1869, 2010.
[14] D. Ehrlich, H. D. Guo, K. Molch, J. W. Ma, and M. Pesaresi, “Identifying damage caused by the 2008 Wenchuan earthquake from VHR remote sensing data,” International Journal of Digital Earth, vol. 2, pp. 309-326, 2009.
[15] G. Lemoine, “Validation of building damage assessments based on post-Haiti 2010 earthquake imagery using multi-source reference data,” in VALgEO 2010 – 2nd International workshop on Validation of geo-information products for crisis management, Ispra, Italy, 2010.
[16] J. Q. Zhang, J. H. Gong, H. Lin, G. Wang, J. L. Huang, J. Zhu, B. L. Xu, and J. Teng, “Design and development of Distributed Virtual Geographic Environment system based on web services,” Information Sciences, vol. 177, pp. 3968-3980, Oct 2007.
[17] A. J. Clark, P. Holliday, R. Chau, H. Eisenberg, and M. Chau, “Collaborative geospatial data as applied to disaster relief: Haiti 2010,” in Security Technology, Disaster Recovery and Business Continuity. vol. 122, T.-H. Kim, W.-C. Fang, and M. K. Khan, Eds., ed: Springer Berlin Heidelberg, 2010, pp. 250-258.
[18] L. Maiyo, N. Kerle, and B. Kobben, “Collaborative post-disaster damage mapping via Geo Web Services,” in Cartography and Geoinformatics for Early Warning and Emergency Management: Towards Better Solutions. vol. 2, M. Konecny, S. Zlatanova, T. Bandrova, and L. Friedmannova, Eds., ed: Springer, 2010, pp. 221-231.
[19] A. M. MacEachren and G. R. Cai, “Supporting group work in crisis management: visually mediated human-GIS-human dialogue,” Environment and Planning B-Planning & Design, vol. 33, pp. 435-456, May 2006.
Dr. Norman Kerle, received Masters degrees in geography from the Ohio State University, Columbus, in 1997 and the University of Hamburg, Hamburg, Germany, in 1998 and a Ph.D. in geography (volcano remote sensing) from the University of Cambridge, Cambridge, U.K., in 2002. He is currently an Assistant Professor in disaster geoinformation management with the Department of Earth Systems Analysis, Faculty of Geo-Information Science and Earth Observation (ITC), Twente University, Enschede (The Netherlands). His research interests include imaging properties and operation characteristics of optical sensors, and in particular object-oriented image analysis (OOA), focusing on hazards and risks, and post disaster damage mapping (sse