Mapping Forest Fire Trajectory using GeoDjango

The project was based on research by the Rocky Mountain Research Station, part of the U.S. Forest Service. The client company had compiled data and some code, but a website was needed to both display the maps and feed information to the planned simulation programs.

In fighting a forest fire and determining where to use resources, there are a lot of variables to consider. The web service needed to account for regional weather patterns and geographic features such as streams and roads in order to compute a Suppression Difficulty Index (how difficult it is to stop a fire given a certain origin point) map for deriving Potential Control Locations (PCLs) that evolve with the situation on the ground.

Four Six Feet Up developers working from four different countries collaborated with two developers from the data service firm on the project, which came down to two major parts:

• a backend simulation program that creates a Suppression Difficulty Index (SDI) for fires given a particular origin point; and
• a website that compiles data on ongoing fires, catalogs them, and generates maps with PCLs based on simulations run with the real world data.

Forest fire simulations

Six Feet Up developers created a service to drive burn simulations using data and formulas from the research. The team used Python as much as possible to build on strands of code that had been developed prior to Six Feet Up’s involvement. Developers used Flask to wrap the simulations and Wine to expose Windows applications as a web service on Linux.

Acknowledging that it could be difficult to get heavy machinery into some areas, the simulations account for the landscape when determining PCLs, and project the fire’s trajectory over a period of two weeks, accounting for weather conditions and topography.

Building the website

The website was constructed 100% by Six Feet Up’s team of experts using GeoDjango to deliver maps with PCLs determined using the simulations. It collects information and documents entered by the U.S. Forest Service and FEMA about wildfires. These documents include technical information about the fires such as the estimated origin point, which is then used by the simulations to determine projections, an SDI, and PCLs for GeoDjango to map out.

GeoDjango handles mapping and coordinate systems, as well as projections between differing coordinate systems. That was particularly useful as not all parties involved in the project used the same type of coordinate systems. If incompatible coordinates are used and GeoDjango cannot convert the coordinates to the proper coordinate system, it sends an error message rather than incorporating the coordinates, which helps keep the data used in the web service clean.

Future work on the project includes a planned website that would run simulations and create a map with PCLs. Recommended elements of the site include:

• areas to dig a fire containment trench;
• controlled burn areas; and
• aerial drop points for aircraft carrying water.

In addition to PCLs, maps generated by the website could include:

• water sources;
• trails; and
• high-value resources to protect.

Six Feet Up expert developers contributed code that will help:

• identify where wildfires are expected to spread, including the most dangerous — and the most safe — locations for firefighters on the frontlines;
• minimize the damage of wildfires by helping constrain the blaze; and
• protect the lives, homes and properties impacted by wildfires by using resources more efficiently.

Six Feet Up’s work laid the groundwork for future complex, predictive-technology projects. Lessons learned from the project can also help other companies productionize machine learning efforts by pairing subject matter experts with the right team of developers. A similar dynamic was employed when Six Feet Up teamed with FLASH to leverage Django and AWS to predict lightning strikes.

Building on this work can lead to even more groundbreaking machine learning (ML) efforts. For example, what if this technology could not only recommend PCLs, but predict the fires in the first place? Read more of Six Feet Up’s ML/AI case studies and explore the company’s ML/AI capabilities.