Worldwide scientific community assists in relief, response and PDNA activities
In the hours following the January 12, 2010 Haiti earthquake, it quickly became evident that this was no ordinary disaster. Multi-lateral organizations were clamoring to discover the true picture of damage to the area and its effect on the people of Haiti, to guide relief, to expedite aid and lay the ground work for the Post-Disaster Needs Assessment (PNDA). One such body, The World Bank, turned to California- and London-based company ImageCat for help. Within 48 hours, GEO-CAN (a worldwide network of scientists) had been activated, and the first maps showing collapsed buildings were being developed. Yet this was only the beginning of a unique response effort from the remote sensing community.
John Bevington Senior Project Scientist, ImageCat UK Beverley Adams Managing Director, ImageCat UK Ron Eguchi President and CEO, ImageCat www.imagecatinc.com
Gathering damage information from remotely sensed data is not a new phenomenon and has been the mainstay of the scientific community for well over a decade. International initiatives are firmly established, allowing rapid access to data for affected nations through the International Charter for Space and Major Disasters. The wider scientific community has often relied on sourcing data from private aerial surveys or tasking of commercial satellite sensors. This was the case following the 2008 Sichuan, China, earthquake, where a highly motivated consortium of scient-ists volunteered their time to a prototype community-based method of damage assessment. This resulted in the development by ImageCat (Long Beach, Calif. and Ashtead, Surrey, U.K.) of the online Virtual Disaster Viewer (VDV; www.virtualdisasterviewer.com), an integrated system allowing pre- and post-event very high spatial resolution imagery to be used for damage assessment.
The system incorporates additional information collected by ground field teams for the understanding of natural disasters by a global audience. For China, a damage assessment of the town of Ying Xiu was completed using pre- and post-event imagery from Quickbird satellite sensors, with each building in the imagery classified according to its damage state using the European Macroseismic Scale 1998 (EMS-98) building damage scale.1 The study area was subset into grid cells and allocated to remote volunteers for analysis at a per-building level to great effect. VDV was further customized for the 2008 L’Aquila, Italy, 2009 Samoa and Sumatra earthquakes, and retroactively for the 1999 Izmit, Turkey, earthquake and the 2004 Hurricane Charley, growing in functionality for each event.
After the Haiti event, impact assessment from remote sensing was the most viable option, due to the spatial scale and magnitude of the humanitarian emergency. This magnitude was compounded by access limitations to the country and a lack of central administration due to damaged and destroyed critical infrastructure leading to a lack of immediate security. The World Bank (WB), through the Global Facility for Disaster Reduction and Recovery (GFDRR; Washington D.C.) moved quickly to utilize ImageCat’s pedigree in multi-hazard damage assessment. Previous experience has ranged from the World Trade Center to Hurricanes Charley, Katrina, and Gustav, and to tornados, wildfires, tsunamis and earthquakes).
At once, a wider team was coordinated to include the staff at Rochester Institute of Technology (RIT), New York, who facilitated an aerial survey of Haiti using their aerial reconnaissance system — the Wildfire Airborne Sensor Platform (WASP).2 In tandem, a multi-phase damage assessment using image interpretation was implemented, first from satellite, then from aerial imagery.
Dr. Stuart Gill from the Disaster Risk Management team at World Bank stated at the height of the response effort (January 19), “ImageCat are working tirelessly to collect and produce a product that will be of enormous value to the reconstruction and damage assessment of Haiti. By making the data they collect Public Domain, we hope to significantly expand the utility of this data”.
Phase 1 of the Haiti damage assessment used 50-cm GeoEye-1 imagery, collected the day after the earthquake, and made publicly available by Google. The consortium assembled for the China earthquake damage assessment was quickly mobilized and VDV was customized for Haiti, allowing 500 × 500-meter subsets of the Port-au-Prince area to be assigned to each volunteer. Within 48 hours, a team of over fifty remote sensing scientists identified more than 5,000 completely collapsed buildings (equating to level 5 on the EMS-98 damage scale) across an area of 133 km2.
These data were delivered to WB/GFDRR, and used for response efforts and identification of the most needy areas for aid delivery. These data also acted as a precursor for a more in-depth forensic analysis using aerial imagery collected by partner institution RIT. The WB-GFDRR-ImageCat-RIT team flew an area of 1,025 km2 using the airborne WASP system, a unique and diverse combination of very high spatial resolution (as fine as 15 cm) visible and infra-red sensors, and also 3-dimensional LiDAR data.
The core team of remote sensing volunteers from Phase 1 was swiftly expanded for the second phase damage assessment and formally recognized as the Global Earth Observation Catastrophe Assessment Network (GEO-CAN). The network of volunteers grew to over 600 participants representing 131 private and academic institutions in 23 different countries, from Sudan to China and Germany to Costa Rica. Several strategic organizations were approached to participate in the Phase 2 damage assessment, each with its own established expertise in engineering or remote sensing.
Previous partners and users of VDV volunteered, including the Earthquake Engineering Research Institute (EERI); the U.K.-based Earthquake Engineering Field Investigation Team (EEFIT); MCEER and LESAM at the State University of New York at Buffalo; Cambridge University, U.K.; and University College London, U.K., to name but a few. Each participant member of GEO-CAN was an established professional in his or her field, and was personally coordinated by ImageCat through VDV.
With initial focus of Phase 1 on Haiti’s capital, Port-au-Prince, the study area was expanded for Phase 2 to include both collapsed and heavily damaged (grades 4 and 5, EMS-98) areas of Carrefour, Delmas, Léogâne, Jacmel, Grand Goave and Petit Goave, a greater than seven-fold increase in the area targeted in Phase 1. Analysis of these areas was also more in-depth, with each heavily damaged and completely collapsed building identified and its image footprint digitized into a GIS database of nearly 30,000 polygons. The distributed damage assessment so effective in Phase 1 was repeated, with each volunteer accessing a restricted area of VDV to check out one grid cell at a time for analysis. Post-event imagery consisted of the 15-cm optical aerial imagery collected by the WB-ImageCat-RIT Remote Sensing Mission and Google aerial mission, and these were compared with pre-event satellite imagery made available by Google Earth. Once a volunteer had completed each assignment, each 2,500 m2 cell would be checked in VDV, and a GIS file of collapsed and heavily damaged building footprints submitted to a central repository. Each building was also assigned a confidence level and descriptive comments based on the user’s assurance of the damage state for each building.
Data from the expanded Phase 2 was delivered in stages to WB/GFDRR following a rigorous quality assessment by ImageCat. Over a third of the main affected areas around Port-au-Prince and the city’s suburbs was delivered within 96 hours of commencement of Phase 2, and within a week, close to 30,000 buildings had been identified as heavily damaged or collapsed across the study area, using solely remotely sensed imagery as a data source. The data from this phase was also independently verified using field ground surveys by multi-lateral bodies, including The United Nations Institute for Training and Research (UNITAR)/Operational Satellite Applications Programme (UNOSAT), the European Commission (EC)/Joint Research Centre (JRC), the Centre National d’Information Géo-Spatial (CNIGS) representing the government of Haiti, and other U.S. teams (Stanford University and Betero-Fierro-Perry, Inc.).
An additional engineering team, led by Cambridge University, U.K., conducted more detailed QA evaluations using very high spatial resolution oblique imagery, provided by Pictometry, which was invaluable in determining whether significant structural damage had occurred that did not result in the complete collapse of a structure (e.g., soft-storey collapse). These ventures helped to validate that the assessments produced from the GEO-CAN initiative were accurate in identifying the total number of collapsed structures and that, statistically, the aerial results could be used as an index for estimating damage at all lower levels (i.e., below ‘collapsed’ and ‘very heavy damage’ – grades 1-3, EMS-98).
The building damage data developed by the GEO-CAN community was further analyzed for land-use type and is currently being used in the ongoing PDNA produced by the Haitian government alongside UNITAR/UNOSAT, EC/JRC, WB/GFDRR and CNIGS. The building footprint information has been used to calculate floor space and an estimated cost for reconstruction and recovery of all collapsed and heavily damaged buildings, approximated by a joint UNOSAT-JRC-WB/ImageCat report to be over $6 billion (U.S.). All data produced in the project have been made publicly available through the World Bank and can be viewed in VDV, alongside ancillary vector data and hundreds of geo-tagged photographs from several post-disaster ground field teams.
The global response to the Haiti earthquake was remarkable in many ways. Data from different missions—World Bank-ImageCat-RIT Remote Sensing Mission (15-cm optical and 2-pt/m2 LiDAR), Google (15-cm optical), NOAA (25-cm optical), and Pictometry, as well as satellite imagery from GeoEye and DigitalGlobe—have allowed damage from the Haiti earthquake to be viewed through multiple sensors and at different times.
Within days, the most recently collected satellite and aerial imagery was publicly available within Google, Yahoo! and Bing Websites, and by the end of January, anyone could access up to half a dozen post-event images for any one area around greater Port-au-Prince.
What has emerged from the Haiti earthquake is truly a new paradigm for global disaster response.
— Ron Eguchi
It was through these international understandings and the dedication of time and resources by GEO-CAN affiliates that a comprehensive and scientifically rigorous damage assessment was achieved. Social networking (through sites such as Twitter, Facebook and Ushahidi) and crowd-sourcing (most notably by the OpenStreetMap initiative) were widely used in the initial response to the Haitian earthquake, and they were effectively used in this damage assessment initiative by GEO-CAN, allowing a comprehensive damage assessment to be achieved remotely using participatory technologies such as VDV and Google Earth.
“GEO-CAN represents the overwhelming response from the international academic community to assist in the damage assessment and reconstruction of Haiti,” said Saroj K. Jha, Manager and Head of the GFDRR Secretariat. “We are very pleased to work with ImageCat in leveraging the expertise of engineers and scientists throughout the world to quickly and accurately assess damage to buildings and critical infrastructure in Port-au-Prince and other affected areas.”
There is huge potential for the international community to leverage GEO-CAN for rapid damage analysis for buildings, infrastructure and identification of temporary relief camps following devastating natural disasters. Additional analysis for Haiti is ongoing by a small section of GEO-CAN, focusing on the identification of areas affected by liquefaction around Port-au-Prince, using the distributed analysis methodology implemented for building damage. At the time of this writing, the GEO-CAN community is gearing up for a building damage assessment similar to that of Haiti for the February 27, 2010 Chile earthquake, and assessments will be deployed when needed by international bodies following future events. Until then, the community continues to grow, and scientists are registering their interest in future participations through the ImageCat website (www.imagecatinc.com).
GEO-CAN provides the mechanism for engaging the disaster management community and exploiting their wealth of combined knowledge in an open and participatory way. Ron Eguchi, CEO of ImageCat, Long Beach, sums up the initiative: “What we’ve learned from the Haiti earthquake is that engineers and scientists, if given the chance, would jump at the opportunity to help after a major disaster. In the past, helping meant traveling to the affected area where only a handful of experts could directly contribute. With GEO-CAN, experts from all around the world can play a direct role in the response by analyzing high-resolution imagery for post-event damage and by contributing their advice and expertise through a social networking platform. What has emerged from the Haiti earthquake is truly a new paradigm for global disaster response.”