On the tail end of one of the hottest summers on record, the dangers of structure fires and wildfires in cities across the United States are continuing to spread. According to NASA, summer temperatures are predicted to continue rising due to climate change, exacerbating heat waves in the western and central parts of the United States in the summer months.
In search of more efficient solutions to the growing problem, the Los Angeles and Louisville fire departments are turning to big data and analysis to address wildfire containment and fire safety, respectively.
The fire department in Los Angeles is working on improving its responses to fires with a new program called FireMap developed by WiFire Lab, a company that develops predictive systems for natural hazards monitoring, simulation, and response by using data-driven models, in San Diego. The program makes predictions about where wildfires will spread next based on real-time information about weather conditions, topography, on-the-ground sensors, government datasets, and flammable material, among other factors.
“When we say ‘fight fires,’ it’s as much about the actual response as it is about the deployment of resources in order to respond. And I think that’s the big piece of where we’re using data—it’s to get better at how we manage the resources that we have,” said Scott Porter, the chief information officer of the Los Angeles Fire Department (LAFD). “We have a lot of institutional knowledge and we make most of our deployment decisions based on people’s experience, but that can be inconsistent and it can also be misleading.”
When firefighters are dispatched to an emergency, making fast decisions in the initial action phase is key to their response success. “There’s a need in the initial phase of a fire to start figuring out: what is the rate of this fire going to be? How do I begin to predict that? What is a realistic containment area for the fire? Where is the best place to position resources in relationship to that? What communities do we need to give early warning notification to?” explained Carlos Calvillo, LAFD’s fire assistant chief.
The predictive model that the dispatch center creates using FireMap can now be pulled up by an incident commander at the scene of a fire on tablets or monitors in their fire truck. The predictive model includes different layerings of views to choose from: satellite imagery, topographic view, or street view where firefighters can identify which streets need to be evacuated. Other layers allow the firefighters to visualize the utility lines in the city in order to decipher whether electrical transmission or gas lines are running through the path of the fire and coordinate with their partners and electrical power supply liaisons at the command center as assisting agencies.
For an incident commander, this information helps them decide what communities they should concentrate their evacuation efforts on, where to place firefighters at the scene, and where to concentrate their efforts. Furthermore, the FireMap program allows command and dispatch centers to more easily communicate and send data to fire chiefs and incident commanders on their computers in the field when responding to a fire.
“What we’re doing mostly today is trying to provide better information in a more timely way to managers and decision makers. So the way we’re integrating big data is we’re making it available to them reports and dashboards that are easier for them to access,” said Porter of the technology firefighters in the field are using. “Last year we started a project to create dashboards that are accessible from the fire stations, from the bureau offices so everyone’s looking at the same information.”
In the past, different fire stations in the city would produce their own spreadsheets and their own reports and there would be conflicts of how they came to conclusions looking at their data. The WiFire program centralized that data across four operational bureaus and 106 fire stations in Los Angeles County.
Now in its third year of usage by the LAFD, FireMap, which began as a prototype program with the WiFire Lab in 2016, has become increasingly valuable to the department in bringing a large department responsible for approximately 4 million people who live in the county’s jurisdiction into the same information source. The tool, created by WiFire researchers, performs data-driven predictive modeling and analysis of fires that have a high potential for rapid spread, and enables predictive analysis of fire scenarios ahead of the time in addition to real-time fire forecasting. The tool also provides easy access to information on past fires, past and current weather conditions, satellite detections, and information on vegetation and landscapes from a multitude of sources. These are all from datasets available on different websites that viewers can now see in one place and be used for planning fire response and management of natural resources ahead of time.
“Putting data in the hands of users is actually very powerful because it’s not just a data scientist looking at it—it’s the actual user and they’re trying to weigh out what the data is telling them versus what their experience is telling them,” said Porter. “ Getting on the same page in terms of a centralized source—one source of truth—but then providing access to the data has really been powerful.”
Calvillo added that the program is also beneficial in training firefighters for future wildfire emergencies. “Pre-incident, I want to get my other captains together to start to sit with each other and start to think about ‘Ok, where is this fire going to go if it starts here?’” he said. “And to actually have that visual and start thinking about ‘Where are we going to place companies now? Where is my evacuation shelter? Where are my evacuation routes?’ To really walk through all of that on a table top—that becomes extremely valuable and that’s another great use of the tool.”
In Louisville, Data Officer Michael Schnuerle partnered with grad students at the Master of Urban Spatial Analytics (MUSA) program at UPenn this spring to work on a city data project with the fire department. Aimed towards fire safety, the project analyzed different datasets to come up with a solution for how the Louisville Fire Department (LFD) could optimize smoke detector outreach to achieve federal requirements and certification.
“It seemed like a really good project to take all the information that we already had around fire incidents, inspection schedules, outreach programs ,and vacant property fires and basically throw all that at the University of Pennsylvania program and see what they could come up with,” said Schnuerle.
From a number of different datasets provided by the city’s property evaluation administrator, the UPenn researchers determined high fire risk factors from three different categories: building risk factors (materials used in wall construction, year the structure was built, density of buildings and vacant structures), neighborhood risk factors (electrical and HVAC permits, distance to vacant or abandoned properties), and demographic risk factors (household income, total population of the area, median rent).
The result was a data project that created a geospatial fire risk index in the Louisville metropolitan area, and then used the calculated risk factors to help inform the fire department in its smoke detector outreach program efforts.
Previously, city employees would go door to door to talk to residents, which wasn’t time effective, or they would visit to neighborhood schools, libraries, or community centers to distribute free smoke detectors and tell residents how to install it.
“The problem with that was they were not really sure if they were reaching the people who would then go back to a house that was in a high risk area or not,” said Schnuerle. Louisville city employees didn’t have a way to track or prioritize which neighborhoods they went to and made decisions solely based on the quantity of building fires that occurred, which was not useful, he explained.
“What they’re doing instead is they’re prioritizing the highest risk areas in each fire district,” said Schnuerle. “And for each district, taking the top two hundred actual street address properties that had the highest risk and making sure to visit those in person. That way they feel better about reaching the appropriate households that are in a high risk area based on this analysis.”
The hope for the future of the project is that in a year or two, city officials will be able to look at the data for fires that occurred and actual fire damage to assess whether there was a change in areas where smoke detectors were distributed versus areas they were not distributed the year before to see what improvements were made, if any. Theoretically, Schneurle explained, if there’s a smoke detector in a building, residents will get alerted and they will be able to evacuate and alert the fire department more quickly, reducing damage and injuries.
But for now, the researchers are working on ways to make this data useful for Louisville firefighters. With the fire risk index data, Schnuerle and the UPenn grad students created a geospatial join over the fire districts and every street address in the city to merge that data with the district so that every street address had a risk associated with it and a fire district assigned. They then extracted the data into a spreadsheet so that it can be easily filtered by the fire department, either by district, risk,or another factor.
“The next step, which is what they’re working on now, is actually integrating some of that data into the vendor system that used to manage inspections and properties,” said Schnuerle. “So essentially, for each address they are adding that information in and working with the vendor so it’s in the system they use every day.” Vendors will also be able to extract data from that system, so they’ll be able to later pull up the scores again along with the address, number of fires, and damage done per address when conducting inspections.
For Schnuerle, as Louisville’s data officer, collaborative projects like this brings together city employees and local graduate students to solve civic problems in an effective way. “This partnership with the university that’s doing this machine learning and analysis and creating a product in conjunction with the city is really a useful model for cities to use because we would not have had the time or the skill set to do this analysis on our own in a timely way,” he said. “So by partnering with a university, you can do something like this collaboratively and it’s just a win for everybody all around—for the residents, for the fire department, for the university, for the students.”