Mapping wildland fire risk using machine learning

Faculty Mentors:

  • Brian Reich (Statistics, NCSU)
  • Margaret Johnson (NASA Jet Propulsion Laboratory)

Prerequisites: Linear algebra, machine learning, programming

Outline: The Western United States is experiencing one of the worst droughts in centuries.  The drought has created the conditions for unprecedented wildland fires that have grave health and environmental impacts. Decision makers need highly resolved spatial and temporal information to mitigate wildland fire impacts. Towards this end, NASA [1] recently established the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) satellite mission to monitor water stress in plants at high spatial resolution. ECOSTRESS’s data provide new opportunities to quantify the links between water use/drought stress and wildfire risk.

Objectives: Students will work with NASA scientists to build a predictive model for the onset and magnitude of wildland fires.  The team will collate water stress images with other geospatial variables and use machine learning to build a predictive model of fire occurrence and severity.

Outcomes: This project will advance our understanding of the conditions conducive to fire and provide managers with valuable information to design interventions strategies.

References:

[1] Pascolini-Campbell, M., et al., ECOSTRESS reveals pre-fire vegetation controls on burn severity for Southern California wildfires of 2020. Glob Ecol Biogeog, 2022. 31: p. 1976-1989.