Year of Award


Document Type


Degree Type

Master of Science (MS)

Degree Name


Department or School/College

College of Forestry and Conservation

Committee Chair

Woodam Chung

Commitee Members

Carl Seielstad, Jonathan Graham


fire occurrence, fire spread, spatial point processing, forest fire risk assessment


University of Montana

Subject Categories

Forest Management


Forest fire risk assessment becomes critical for developing forest and fire management strategies in Korea since the magnitude of damage from fires significantly increased over the past decades. Fire behavior probability is one of the major components in quantifying fire risk, and is often presented as burn probability. Burn probability estimation requires a proper estimation of fire occurrence probability because fire spread is largely influenced by ignition locations in addition to other environmental factors, such as weather, topography, and land covers.

The objective of this study is to assess forest fire risk over a large forested landscape in and around the City of Gyeongju, Republic of Korea, while incorporating fire occurrence probability into estimation of burn probability. A fire occurrence probability model with spatial covariates and autocorrelation was developed using historical record of fire occurrence between 1991 and 2012 and a spatial point processing (SPP) method. A total of 502 fire ignition points were generated using the fire occurrence probability model. Monte Carlo fire spread simulations were performed from the ignition points under the 95% extreme weather scenario, resulting in burn probability estimation for each land parcel across the landscape. Finally, the burn probability was combined with government-appraised land property value to assess potential loss value per land parcel due to forest fires.

The density of forest fires of the study landscape was associated with lower elevation, moderate slope, coniferous land cover, distance to road, and higher tomb density. A positive spatial autocorrelations between the locations of fire ignition was also found. An area-interaction point process model including the spatial covariate effects and interpoint interaction term appeared to be suitable as a fire occurrence probability model. A correlation analysis among the fire occurrence probability, burn probability, land property value, and potential value loss indicates that fire risk is largely associated with spatial pattern of burn probability (Pearson’s correlation =0.7084). These results can provide forest and fire management authorities in the study region with useful information for decision making. It is also hoped that the methodology presented here can provide an improved framework for assessing fire risk in other regions.



© Copyright 2015 Hyeyoung Woo