Applications of space-time point processes in wildfire forecasting

Rick Schoenberg
University of California, Los Angeles (UCLA)

The Burning Index (BI) is commonly used as a predictor of
wildfire activity. An examination of data on the BI and wildfires in
Los Angeles County, California from January 1976 to December 2000
reveals that although the BI is positively associated with wildfire
occurrence, its predictive value is quite limited. Wind speed alone
has a higher correlation with burn area than BI, for instance. The
use of
alternative point process models will be explored, and in particular
it will be shown that a simple point process model using wind speed,
relative humidity, precipitation and temperature well outperforms the
BI in terms of predictive power.

Audio (MP3 File, Podcast Ready) Presentation (PowerPoint File)

Back to Crime Hot Spots: Behavioral, Computational and Mathematical Models