FireCast is an algorithm integrated in a program called the Risked Based Inspection System (RBIS) and in used in New York since 2013 to compute the risks of a fire in a building and create lists of the properties to inspect in priority.
The version 2.0 of this system gathers data from 5 city agencies, and uses 60 fire risks factors to assess the risk for each of the 330’000 buildings of the city.(http://www.nfpa.
org/news-and-research/publications/nfpa-journal/2014/november-december-2014/features/in-pursuit-of-smart ). In contrast to Firebird, FireCast 2.0 is not an open-source framework, hence there is no scientific publication, no code accessible and very little communications about the results.
Since the deployment, 16.5% of the structure fires happenned in buildings that the FDNY inspected within 90 days before the incident ((http://www.nfpa.
org/news-and-research/publications/nfpa-journal/2014/november-december-2014/features/in-pursuit-of-smart), suggesting that the algorithm was able to target the right buildings. The goal for the inspectors of the New York City Fire Department (FDNY) is to visit around 30’000 buildings per year. (https://www.smithsonianmag.
com/innovation/how-data-and-good-algorithm-can-help-predict-where-fires-will-start-180954436/). Before the RBIS was in used, inspections would rely on informations communicated using cards on which basic information about the buildings were written (occupancy, superficy, year built…). It was up to the managers of the fire department to determine the frequency of the inspection for each building. The cards were stored in different base stations and nothing was digitized, making it hard to keep track of the information, and impossible to prioritize the inspections. FireCast and the RBIS automate this workflow and help the management of the fire department meet goals and improve the safety of the citizens and the firefighters.Similarly to Firebird, a vast amount of data has to be collected in order to predict the risks of fires. FireCast uses data from the FDNY, the fire code violations, 311 service requests, the Department of Buildings, weather and private sector agencies.
(https://medium.com/homeland-security/big-data-providing-fire-service-and-emergency-response-communities-with-tools-to-predict-and-fecacba466cc). Here, the difficulty to collect and join the data is not publicly documented, but in a city as large as New York, it can be considered similar to Atlanta.
Furthermore, the new version of the algorithm, FireCast 3.0 is expect to be launched since 2015 (https://www.smithsonianmag.com/innovation/how-data-and-good-algorithm-can-help-predict-where-fires-will-start-180954436/ ) but has yet to be released (https://medium.com/homeland-security/big-data-providing-fire-service-and-emergency-response-communities-with-tools-to-predict-and-fecacba466cc). The new algorithm will use more data (coming from 17 agencies) and more features (up to 7500).
The underlying reasons for the delay is the organizational barriers among the different actors that prevent the data to be easily transmitted, an essential point, as demonstrated by Firebird. The greatest challenge in this framework is once again the organizational reticence to share data and manage an automated workflow not equipped with the human capital suited for this task.Recommendations and roadmaps The researchers involved in the Firebird project suggested revising the municipal fire code so that the inspections type, frequency and priority could rely on data-driven decisions, similarly as in British Columbia where the fire code was changed to rely on data and quantitative metrics rather than tradition and intuition. (https://www.surrey.ca/files/RiskAssessmentToolUFVResearchNoteFullReport.pdf). Moreover, they advised municipalities to begin a “significant rethinking” of how to exploit and share data among agencies as for example the creation a common database containing a unique identifier for each building.
Finally, as they showed the barriers that exist with the adoption of these new predictive tools, they advised to conduct more researches on the usability of the interactive user-interface and how the management could use these tools in their day-to-day operations.In 2013, the data analytics office in New York created Databridge, a citywide data-sharing analytics platform joining data from 40 agencies. (https://gcn.com/Articles/2013/10/04/GCN-Award-NYC-DataBridge.aspx ) This innovation was the first stone of the FireCast project as it has enabled the data-sharing process across multiple databases from different actors. For instance, the platform has allowed to locate structures that have been illegally subdivided, thus at risk of catching fire. Databridge is a good example to show that it is possible for different actors with distinct missions to join their data in a structured centralized platform, so that they can use each other’s databases for different purposes.However, with such large systems, the data has to be uniformized.
Furthermore, if data from fire departments was uniformized, development in smart firefighting could be accelerated. It was from this perspective that the NFPA released the norm NFPA 951, Guide to Building and Utilizing Digital Information (http://www.nfpa.org/codes-and-standards/all-codes-and-standards/list-of-codes-and-standards/detail?code=951 ) and the norm NFPA 950, Data Development and Exchange for the Fire Service. (http://www.nfpa.org/codes-and-standards/all-codes-and-standards/list-of-codes-and-standards/detail?code=950 )