Predicting Locations of Emergency & Damage During Disaster Using VGI Data

[Guest blog post by Prateek Budhwar who is currently pursuing M.Tech from Indian Institute of Technology, Roorkee, India. Prateek is interested in the areas of Volunteered geographic information, Digital image processing, GPS based LBS applications on android and remote sensing]. The VGI data obtained from the Ushahidi-Haiti platform during the first 72 hours of Earthquake in Haiti can be used for predicting the locations of ‘Emergency and damaged areas’ using Ushahidi Reports in Port-au-Prince. Two rapid and very successful VGI deployments helped coordinate disaster response after a devastating magnitude 7 earthquake struck Haiti in January 2010. OpenStreetMap (OSM) project volunteers working outside Haiti created a digital street map of Port-au-Prince and other places in Haiti very rapidly using fine-resolution imagery to trace vector maps of streets and other features. The Ushahidi Project was able to post appeals for help, translated from Creole into English by another group of online volunteers. Together, these VGI projects were instrumental in guiding first responders to disaster victims. Numerous case studies stressed the added value of using VGI in various types of crisis events, such as earthquakes (De Rubeis et al. 2009), forest fires (De Longueville et al. 2009), political crises, hurricanes (Hughes and Palen 2009), floods (De Longueville et al. 2010a), and terrorist attacks (Palen et al. 2009) which is not possible using the GIS. Current mobile geographic computing integrates four key technologies that were previously separate systems. Widely available mobile devices now include GPS, basic GIS software, wireless communication, and access to the Internet - all in a handheld or tablet computer (Feick R. 2010). So we can get reports on incidents and damages in the area affected by the earthquake from the locals to provide immediate help. The Ushahidi Haiti Project (UHP) was a volunteer-driven effort to produce a earthquake crisis map after the January 12, 2010 earthquake in Haiti. The project represents an impressive proof of concept for the application of crisis mapping and crowdsourcing to large scale catastrophe and a novel approach to the rapidly evolving field of crisis informatics. Steps for predicting the locations of damages and emergencies from the code by  Drew Conway: 1.    Get the VGI data in form of CSV (Comma Separated Value) format file from the Ushahidi platform. 2.    Install the R software. 3.    Load libraries geoR and geoRglm in the R software required for the analysis. 4.    Load Ushahidi data. 5.    Create a vector data for general report type, currently there are too many sub-categories like response, water contamination, security. 6.    Create subset of data for values occurring in and around Port-au-Prince. 7.    Create count vectors giving the number of incidents for report types and consolidate duplicates. 8.    Now we will have columns for every event type. 9.    Create geodata object specifying spatial model of the VGI data. 10.  Get spatial weights among incident reports. 11.  Perform Markov Chain Monte Carlo simulation on the VGI data. 12.  Create grid over the Port-au-Prince. 13.  Generate probabilities for Category 1 reports in each grid location. 14.  Plot choropleth of Probability for Earthquake Damage Emergency in Port-au-Prince. RESULT AND OBSERVATION : Result: The result obtained from the analysis is as shown below: Figure 1 : Plot of the probabilities generated for emergency and damage after the Haiti Earthquake from the VGI data The result is obtained by plotting the probability of events belonging to the category 1 i.e. emergency and damage corresponding to the longitude and latitude on the graph which shows the most of the reports coming from near the epicenter of the earthquake i.e. 18.4514ºN  72.4452ºW. Figure 2 : Choropleth of probabilities generated for the category of emergency and damage from the VGI data In the above figure plotted using the heat map, the darker regions are showing more probability of emergency and damage and damage than the brighter regions,  the green mark is representing the epicenter of the earthquake and the blue mark is representing the Port-au-Prince city. Observation: According to the plot obtained from the probability distribution of the VGI data reports from Ushahidi platform, it is observed that the areas which are worst affected by the earthquake and are damaged are giving more reports than the areas which are less affected by the earthquake. So the result obtained from the analysis within the few hours after the earthquake is giving some idea about the crisis dynamics on the ground. The VGI data obtained from the volunteers on the ground can be effectively used by the emergency responders during the disaster management. The result shown in the choropleth are also similar to the damage assesment map prepared by the European Commission in the figure 3. Figure 3 : Haiti Earthquake damage assessment Map CONCLUSION AND RECOMMENDATION Conclusion: The temporal dimension in VGI is important to create an accurate picture of what's occurring where and when in a particular situation. Maps have traditionally emphasized the static aspects of geography. Examining the issue of the quality of VGI, it can readily claim to be the most current data source, as witnessed by Twitter's ability to break news before any other sources. VGI, compared to traditional authoritative data sources, has been criticized for having poorer positional accuracy and overall veracity, however, that assertion is countered by citing Linus's Law, which states that the more people involved and watching over an activity, the more likely errors can be spotted and fixed quickly. The Ushahidi platform collects and stores data in a standard format to  ensure  that  clear   and sufficiently understandable data is available for analysis. After the analysis of the data from Ushahidi reports, it is important that the availability of relevant information is critical. Generally, for a crisis management system, it can be an issue as sometimes the data might be too little and in others it might be too much. The crisis management platform is a collaboration between a lot of organizations hence it is important to isolate the independent contribution of individual organizations. A crisis management system demands rapid deployment after a crisis has occurred. This fact makes it different from the usual planned IT project implementation. The Ushahidi Haiti project was quickly deployed in response to the sudden earthquake in Haiti and thus it presents challenges pertaining to data collection and proper documentation. These factors cause further difficulties for proper evaluation of a crisis management system. There is a lack of relevant theoretical approaches guiding analysis of a crisis management system. It is important to recognize that the realm of crisis management and its analysis is still a novel domain. As the discipline evolves it will become increasingly important to understand, analyze and measure various aspects of it and many other kind of critical information can be extracted from the VGI data during disaster. Recommendation: Improvement in the accuracy of classification approaches can be utilized in crisis mapping through closer collaboration with seasoned field operators, better trained/supervised volunteers, and improved integration of intelligent summary tools with crisis mapping. It should be ensured that short codes and reporting channels/instructions are unambiguous and clear in purpose and use. Improvement in the quality of aggregation of information through better categorization and more intensive use of analytic/visualization tools should be done. Ushahidi should continue to work on improving the capacity of crisis mappers in the area of geolocation, including appropriate level of precision based on need or phase of the crisis to the extent that incident status tracking and updating can be streamlined and adopted by Ushahidi users, impact as well as evidence of impact will also likely be strengthened. New ways should be explored for VGI data to motivate and facilitate users of Ushahidi and similar platforms to “track” and “close” reports. Improve sorting and monitoring of comments and incident status updates. Tighter integration of the Ushahidi web application with major social networks should be considered to help jump start broad user community activity for new implementations. Data structure and processing of reports must be improved to include mandatory metadata to meet an international standard. Information utility can be improved by increasing the diversity of intelligent summary tools and reporting features of the Ushahidi web application. Collection of up-dated analysis summaries in situation reports should be considered, organized in ways that partners can use them and distributed in ways that they can receive them.