Analysis of the seasonal characteristics of forest fires in South Korea using the multivariate analysis approach

Abstract

For efficient forest fire management, special precautions are required in dry and strong-wind seasons vulnerable to severe forest fires. To extract the seasonal characteristics of forest fires in South Korea, the statistics over the past 16 years, 1991 through 2005, were investigated. The daily records of the number of fire occurrences, the total area burned and the average burned area per occurrence were examined to identify the seasonal patterns of forest fires using cluster analysis and principal component analysis; the risk of daily fires was also assessed using the ordered logit model. As a result, the fire patterns were classified into five clusters and a general danger index for forest fires was derived from the first principal component, showing relatively large-scaled fire regimes in spring, and frequent small-scaled fire regimes in autumn and winter. In connection with the ordered logit model, the probability for the five ranks of forest fire risk was calculated and the threshold for high-risk fires was detected. As an implementation of the results above, the proper forest fire precautionary period in South Korea was estimated, and consequently October 21 through May 17 was recognized as a dry season at a high risk of forest fires. This period began 10 days earlier in autumn and extended into midwinter (late December and January) as opposed to the existing precautionary period, indicating the need of more cautious forest fire management earlier in autumn and continuing through midwinter.

Keywords:

• Fire statistics
• Forest fire precautionary period
• Forest fire seasonality
• Statistical analysis