A191-08
Forecasting Fire Risk With Machine Learning And Dynamic Information Derived From Satellite Vegetation Index Time-Series

Tuesday, 15 December 2020: 05:58
Virtual
Yaron Michael1, David Helman2, Oren Glickman3, David Gabay3, Steve Brenner1 and Itamar Lensky1, (1)Bar-Ilan University, Geography and the Environment, Ramat-Gan, Israel, (2)Hebrew University of Jerusalem, Soil and Water Sciences, The Robert H. Smith Faculty of Agriculture, Food and Environment, Rehovot, Israel, (3)Bar-Ilan University, The Data Science Institute, Ramat-Gan, Israel
Recorded Presentation
Abstract:
Fire risk mapping is essential for pre-fire management as well as for efficient firefighting efforts. Most fire risk maps are generated using static information on variables such as topography, vegetation density, and instantaneous wetness status of the fuel matter. Satellites are often used to provide such information. However, long-term vegetation dynamics and the cumulative dryness status of the woody vegetation, which may affect fire occurrence and spread, are rarely considered in fire risk mapping. In this work, we investigate the impact of using two satellite-derived metrics that represent long-term vegetation status and dynamics – NDVIW and NDVIT on fire risk mapping. NDVIW, which is the average value of the normalized difference vegetation index (NDVI) of the woody vegetation, represents the mean woody density at the grid cell. The NDVIT is the 5-year trend of the woody vegetation NDVI in the grid cell. NDVIT represents the woody vegetation long-term dryness status. To produce these metrics, we decompose time-series of satellite-derived NDVI following a previously suggested method adjusted for Mediterranean woodlands and forests. We tested whether these metrics improve fire risk mapping using three machine learning algorithms (Logistic Regression, Random Forest, and XGBoost). We chose the 2007 wildfires in Greece for the analysis. Our results indicate that XGBoost, which accounts for variable interactions and non-linear effects, was the machine learning model that produced the best results. NDVIW improved the model performance, while NDVIT was significant only when NDVIW was high. This NDVIW–NDVIT interaction means that the long-term dryness effect is meaningful only in places of dense woody vegetation. The proposed method can produce more accurate fire risk maps than conventional methods and can supply important dynamic information that may be used in fire behavior models.
<< Previous Abstract | Next Abstract >>