Extreme value analysis of forest fires from New York to Nova Scotia, 1950–2010

Northeastern North America is a region of low fire occurrence compared to other areas. Moreover, reported area burned and fire numbers have been declining to unusually low levels over several decades. Reported forest fire data for 1950-2010 were analyzed for New York, New England, New Brunswick, and Nova Scotia. Several diagnostics for extreme value behavior are described. Statistical analysis showsthat extreme value behavior can be preserved even at lower average levels of fire occurrence. A small number of years accounts for a significant share of cumulative area burned. Since 1957, there have been no years in which area burned exceeded 2 standard deviations from the mean in two or more jurisdictions. Datasets on individual fires since the 1970’s are available for several states. An exploratory analysis of such data for New York shows that skewness in fire size distributions is much more marked for individual fires than for annual totals. Such data permits analysis of peak loads on fire control services in terms of monthly and daily fire activity. Considering the potential risks to life and property, further analysis of extreme fire occurrences in the region is warranted.

► This paper presents a dataset for forest fire history northeastern North America for 1950–2010.
► Annual area burned has declined since the 1950s, but extreme behavior persists.
► A number of useful diagnostics for extreme value behavior are posed.
► Statistical diagnostics for individual fire data are illustrated for New York.
► Averages are misleading for assessing fire histories and planning for fire control.