The Carr fire was my second trip to a disaster zone in California as a data analyst supporting public agency UAV teams. In both this fire and the Tubb’s fire in Santa Rosa last October, my role has been to coordinate drone data capture and mapping products to the right people across the disaster management chain. Essentially, I act as a link between the pilots and those responsible for assessing and rebuilding their community.
The Alameda and Contra Costa County Sheriff’s office, along with Menlo Park Fire Department and San Francisco police arrived in Redding after the Carr fire had burned nearly ~ 1000 homes. A half a dozen lives had already been lost and the fire continued to spread more broadly into rural areas of Shasta County.
Residents had yet to return to many of the evacuated neighbourhoods and, understandably, there was a lot of anxiety for information on what remained of their communities. Our goal was to provide that information quickly.
Teams assembled the first morning at Redding City Hall and the Cal Fire air boss briefed the group. Altitudes were capped at 100-200 feet above ground level, depending on the location within the Temporary Flight Restriction (TFR) zone. If manned aircraft entered an area, drones were to be landed quickly.
Redding police then identified target neighbourhoods where imagery would be most valuable. The city was divided into 5 zones. One drone team was assigned to each zone.
After the briefing came a rapid data onboarding process. Being that we had experienced drone pilots, the focus was on standardizing a protocol for autonomous data capture and streamlining the workflow for different types of drones, apps and imagery outputs.
The drones were all off-the-shelf DJI drones, primarily Phantom 4s and Mavic Pros. Standardizing on a well-known platform simplifies the overall process of data collection. Romeo Durscher, Director of Public Safety for DJI, was also on hand to assist.
Initial challenges were in unestablished workflows, primarily revolving around various pieces of software. If someone wasn’t familiar with a software, the apps are all generally easy enough to use that 10 minutes of close instruction ensured basic competence. Best practices were established for pushing data to various sources, and ensuring that issues were dealt with quickly.
A challenge we faced was the being capped at such low altitudes in the TFR greatly restricted traditional drone mapping of flying 75-80% overlap in photos for photogrammetric processing. At 200 ft, a drone could cover about an acre per minute, but target zone consisted of ~ 1000 acres spread over ten times that size.
In the Tubb’s fire, I had found that 360 drone panoramics were simple, powerful tools for quickly visualizing status on the ground. The teams could efficiently cover a large area of on short notice. So, this was the primary method we used and traditional mapping was restricted to a few crucial areas for the sake of time and drone battery life.
From Redding City Hall, six teams dispersed across the city to their designated zones.
Out in the community, I found the devastation of the Carr fire was not as concentrated as the Tubbs fire nine months before. In Santa Rosa, 3000 homes were incinerated in a single night and the Coffey Park neighborhood resembled a war zone. The Carr fire was more dispersed, with smaller sections of any given neighborhood burned. Yet, the sheer scale was enormous, covering well over a hundred square miles at that point, and this was just one of three major fires burning in the state. In fact, the Mendocino complex fires burning to the south were deemed larger the day we arrived.
Of course, it is incredibly sad to see the destruction of homes. Even more so the loss of both pets and people, including two small children about the age of my own daughter. To be honest, this sadness sticks to the soul long after I’m safely home.
In light of the situation, the teams performed amazingly well. We captured over one hundred 360 degree panoramas across the target zones and were able to map core areas of damage.
As teams poured in hot and tired from the field, they brought with them a tsunami of data. SD cards needed to be sorted, flight logs pulled off of tablets, forgotten passwords reset, and software unexpectedly updated. As high-tech as the drones are, it was sharpies and ziplock bags for memory cards that saved the day at this point.
Then the processing began.
With the help of a lot of caffeine, the data started coming together. 360-degree views of high-resolution imagery of burned neighborhoods began populating as pins on a central map. Orthomosaics were stitched together and city property boundaries overlaid.
I didn’t sleep much that night, as I kept waking up to make sure everything was progressing smoothly.
The next morning was another early briefing and a preliminary look at some of the results. It’s helpful to show the teams some fruits of their labor before they head back out into the heat and smoke-filled air.
New zones were established outside the city into different areas of Shasta county. While teams continued to capture more 360 imagery, I began converting the output files into formats that the city and county could use. This typically involves ESRI products and ArcGIS, standard software for most public GIS divisions. I had learned from the Tubb’s fire it was important to think about where and in what format the data were going to live before you even start collecting it.
The Redding GIS team took over from there and did a fantastic job funneling all the pins of 360 locations to their public map of the fire boundaries and evacuation zones with. The entire process took about 48 hours from first flight to the information going public.
If all this sounds smooth…it wasn’t. Wefacedtablets and phones overheating in the sun, apps crashing, minimal cell reception, and slow internet. Drones needed unlocked in the TFR. There were rogue drones. Helicopters entered the airspace regularly. And, of course, fatigue.
Still, the team got it done. We felt that a victory was won for the drone industry. The technology and process was validated once again in an emergency. Moreover, this happened with full consideration of the sensitive nature of UAVs in fire zones of California. Credit goes to Cal Fire and their air boss for looking toward the future of fire management in a state that will continue to burn.
We drove home to the Bay Area exhausted with clothes smelling of smoke. Every SD card was full, every battery was drained. Empty water bottles littered the floor.
While many new lessons were learned, there is only one I feel is critical to share here. That lesson is that the drone is only half the story in these situations. It is only a data capture tool. Without data analysis and visualization, all the the careful planning and execution by pilots will just sit on a hard drive somewhere in a drawer. As agencies build their drone teams, it is vital they think about hiring or training drone analysts. Job descriptions as follows:
Work effectively with teams of pilots from different agencies
Clearly communicate and rapidly train mixed skill levels on best data capture practices
Manage and process large amounts of UAV data and deliver outputs quickly
Coordinate with the disaster management team on data transfer and visualization
Quickly and creatively solve problems (about 25% of the job is on-the-fly technical troubleshooting)
Carry out and validate micro-experiments to test methodologies for the next emergency
Technically support the GIS and agency teams for at least two weeks afterwards
Don’t be an asshole
I am honored to have been part of the team and to do my own small part in assisting those impacted by the fire. This was all volunteer. I didn’t get paid. My thoughts are my own and do not reflect those of the agencies involved. My hope is the imagery is helpful to the residents and in documenting the scope of the tragedy.
I will be building a Rapid Response Drone Data course to scale these and other methods used in this fire.
You can sign up for news on when training become available at https://www.scholarfarms.com/