January 7th marked the beginning of one of the most devastating wildfire events in California’s recent history. A series of fires struck the city of Los Angeles, and within just 24 hours of the first ignition, the city was already facing three major wildfires in Pacific Palisades, Eaton, and Hurst. At that point, two fatalities had been reported, and over 70,000 people had been evacuated.
Wildfires are not uncommon in California; residents are, in many ways, used to them. However, the conditions surrounding this particular event were unique due to the wind patterns. The Santa Ana winds, which typically push air southward, unusually shifted north this time. As a result, the initial spark spread rapidly with the wind. On top of that, last year’s rainy season in the Los Angeles area was above average, allowing vegetation to grow thick and tall. But this was followed by an extended dry spell, leaving the abundant plant life tinder-dry and highly flammable. Combined with low humidity levels, these natural factors contributed to making the Pacific Palisades fire one of the most destructive in the state’s history.
During the unfolding crisis, residents faced a series of urgent challenges. With two more zones ordered to evacuate, three new evacuation warnings issued, heavy traffic congestion, and three road closures along the Pacific Coast Highway, escaping the flames became an ordeal. Many people even abandoned their vehicles in desperation. At the same time, firefighters were struggling to meet the water demands, as water systems in many areas could not keep up.
Those who chose not to evacuate stayed behind to defend their properties, using garden hoses and even pool water to spray down their homes and surrounding vegetation in an attempt to prevent the flames from spreading. Firefighting helicopters were also seen drawing water from swimming pools and nearby reservoirs to help contain the blaze.
Staying behind in an active fire zone is extremely dangerous. Statistics show a growing number of wildfire-related deaths in recent years. So why would anyone choose to risk their life in such circumstances? The answer lies in part in the changing insurance landscape. In the months leading up to the fires, many insurance companies began withdrawing wildfire coverage. Thousands of policy renewals were denied in 2024 for residents in the Palisades and other LA ZIP codes. For those still eligible, premiums have skyrocketed to levels unaffordable for much of the population.
Insurance providers justify this by citing climate change. As they argue, “climate-related extreme weather events will become both more frequent and more violent, resulting in ever-scarcer insurance and ever-higher premiums.” In other words, losses from climate-related disasters are expected to rise dramatically, and securing home insurance that can withstand such risks is becoming increasingly crucial, yet also increasingly out of reach.
Research Question
How can we use Big Data to identify the key environmental and human factors that contribute to the risk of wildfires worldwide so we can help urban planners identify high-risk vulnerable zones and inform communities about their exposure to this phenomena, while also applying machine learning models to detect urban elements that might help to response to wildfires.
Pipeline
Working in three scales
Pipeline
Macroscale
Aa
Mesoscale
Aa
Microscale
There is no doubt in my mind that this generator, pump, pool, sprinkler and fire hose contributed to saving our family home. We ran this equipment for about 10 hours over the course of 2 days. We drained 2 pools. Without this, there would be no more house and possibly no more neighbors. Were we lucky? Extremely. Were we prepared? Yes. Were there things we would have done differently? Absolutely.
User @cali_._love via Instagram
The community of Pacific Palisades felt the obligation to defend their homes using whatever water resources they had available. For those fortunate enough to have a swimming pool, this became their primary means of protection.
Based on this reality, we decided to use machine learning detection models to map the community’s vulnerability to fire, using the presence of swimming pools as an indicator of potential fire-fighting capacity.
To do this, the first step was to train a machine learning model to detect pools using satellite imagery through the Roboflow platform. Once the model was trained, we exported it as an ONNX file using a Python script. This ONNX file was then used within QGIS through the Deepness plugin to run the detection of pools in the Pacific Palisades area. The result was a shapefile containing all detected pools.
This element detection was carried out in areas previously identified as vulnerable at the mesoscale level. Through this exercise, we aimed to assess whether it is possible to rely on swimming pools as a personal water source to fight fires in zones that are vulnerable from both natural and human perspectives.
Toolkit
Video
Conclusion
Mention comments from the jury.
This tool can be improved, etc etc
