Inspiration
This past year has been especially bad for Californian wildfires, and we want to help. We seek to provide a user-friendly and visually appealing fire-prediction model that correctly responds to major shifts in wind, elevation, humidity, and biome. Software like this not only provides a not-to-scale proof of concept for more advanced fire prediction software, but also presents itself in a way that engages the general public and directs them towards resources for learning more about wildfire safety.
What it does
Our model takes wind speed and direction, humidity and an ignition point from the user as inputs. These metrics - and several hard-coded others such as biome, slope and, "greenness" - affect the spread of a fire across a map of California. This spread is based on existing fire-prediction models produced by the USDA Forest Service, BehavePlus 5.0.
How we built it
The back end was built using python while the front end is a standard html/css/js web page that uses a flask server. Data for topography, biome and "greenness" was pulled from released heat maps and infographics, converted into a 2D grid, and stacked on top of each other to form a layer cake of 2D grids. The biomes, humidity, elevation, and wind speed are combined together according to the framework present in BehavePlus to determine the behavior of a fire at any one time. Four major types of wildfire - crown, surface, conditional crown, and torch - can be present based on the conditions, and each will propagate at a different speed. The framework for tweaking how the biomes interact with other variables is included in the software as well.
Challenges we ran into
The model is slow to return multiple outputs, and due to the high number of interlocking systems bugs were constant and presumably still present. The creation of realistic fire behavior within a certain biome not only requires more than 36 hours of research, but also a meticulous tweaking of internal values to produce accurate behavior.
Accomplishments that we are proud of
Completing a full projects with all of requirements we hoped to complete, effectively working with teammates we have not met before, creating a simple yet very satisfying GUI
What we learned
How to use a flask server. How to effectively split up tasks and utilize teammate strengths.
What's next for Up in Smoke
More maps with smaller scopes to predict smaller scale fires, a higher resolution, and more accurate biome modeling are all possible improvements to this project.
Log in or sign up for Devpost to join the conversation.