The purpose of the project is to provide rapid response to natural disasters by developing a situation aware virtual information observatory to extreme events occurring anywhere in the world. The project focuses on methods by which the multiuser virtual world can have a coherent representation of assets, which differ by type and scale, by assimilating the physical and observational data into operational predictive models. It will utilize cloud computing services to integrate these divergent assets for effective real time data assimilation of multimedia data information streaming and meta-tagging into real world socio-physical models, employing unstructured databases that can be queried in this specific virtual world for distributed viewers and decision makers to experience their impact. These activities will interact with the virtual observatory to make more effective decisions in planning event response by providing a scenario querying and testing platform.
The purpose of this project is to provide a rapid response to natural disasters by developing a situation aware Virtual Extreme Event Observatory (VEEO) occurring anywhere in the world. We have implemented elements of a framework for the virtual information observatory that that indirectly monitors natural disasters based on analyzing frequency reporting from social media postings. The system employs a Human Sensor Network (HSN) that collects geolocated data initially from two media sites, Twitter and Instagram. The data are automatically collected on the UMBC/CHMPR ‘bluegrit’computer system and stored in the Couch database for querying to identify extreme event occurrences. The data is then filtered, queried and analyzed employing an open source scalable software system called Elastic Search engine with a REST API that runs ontop of Lucene under the Hadoop file storage and cloud processing system. For a certain class of events, the system responds to the event by invoking high resolution NOAA operational forecast models to predict event related impacts that require topographic data input specifications as well as other model related data. The social media observations are used as boundary forcings and were shown to improve the operational forecast models and the model output analysis provided as decision inputs for emergency responders. The project focuses on methods by which the multiuser virtual world can have a coherent representation of assets, which differ by type and scale, by assimilating the physical and observational data into operational predictive models. For tracking debris arising from the Fukushima tsunami that can reach the west coast, we employed the General NOAA Operational Modeling Environment (GNOME), which is a 2-D Lagrangian particle flow tracing model forced by Wind and Ocean driven currents. GNOME is utilized operationally by the NOAA Office of Response and Restoration for such events as oil spills, ocean debris, and lost ships and humans at sea. Related Lagrangian models such as Hysplit for aerosol tracking and the Slosh model for hurricane surge prediction have also been incorporated in the VEEO. The implementation of these models have been successfully tested during the Deep Water Horizon oils spill in the Gulf of Mexico and Hurricane Sandy landfall in the New York Bight area. Figure 1 shows the debris sighting data observed on the West coast as collected and evaluated by OR&R. As part of the VEEO system, the framework allows for the filtered integration of heterogeneous HSN sources as well as inputs from a variety of model forecasts for providing decision support data in a readily interpreted form. This Rapid project hasdeveloped interactive social media web-based data model forecast display systems that will exchange information with the NOAA operational Debris Program Office. Our CO-I at UCSD developed a real time streaming situation aware display system. Figure 2 shows an example of the UCSD situation aware display importing topographic elevation data for events occurring anywhere on the planet. Figure 3 illustrates how the UCSD system displays geolocated social media data from youtube streamed to the event sites on realistic topography. At UMBC, we have developed AsonMaps - a flexible framework for harvesting, analyzing and visualizing HSN data superimposed on the operational geophysical model forecasts. The framework allows the filtered integration of heterogeneous HSN sources as well as inputs from a variety of model forecasts. Figure 4 shows a screenshot of AsonMaps centered on the greater New Your City area. The overlay indicates different surge heights in different colors. Geolocated tweets are marked with a blue bird icon and Instagram images with a camera icon. On the left hand side are posts with disabled geolocation. Thus, the main project outcome of this award has been to put in place a working Human Sensor Network framework for a VEEO that should be able to respond in near real time in most parts of the planet with improved social media based observational operational analysis for emergency responder decisions for mitigating the human and societal impacts of natural disasters.
